diff --git a/README.md b/README.md
index 8842108..bdd4c5f 100644
--- a/README.md
+++ b/README.md
@@ -31,7 +31,7 @@ Download it and try it out for free! **https://piebald.ai/**
[](https://github.com/hesreallyhim/awesome-claude-code)
-This repository contains an up-to-date list of all Claude Code's various system prompts and their associated token counts as of **[Claude Code v2.1.323](https://www.npmjs.com/package/@anthropic-ai/claude-code/v/2.1.323) (November 12th, 2026).** It also contains a [**CHANGELOG.md**](./CHANGELOG.md) for the system prompts across 335 versions since v2.0.14. From the team behind [
**Piebald.**](https://piebald.ai/)
+This repository contains an up-to-date list of all Claude Code's various system prompts and their associated token counts as of **[Claude Code v2.1.324](https://www.npmjs.com/package/@anthropic-ai/claude-code/v/2.1.324) (November 13th, 2026).** It also contains a [**CHANGELOG.md**](./CHANGELOG.md) for the system prompts across 336 versions since v2.0.14. From the team behind [
**Piebald.**](https://piebald.ai/)
**This repository is updated within minutes of each Claude Code release. See the [changelog](./CHANGELOG.md), and follow [@PiebaldAI](https://x.com/PiebaldAI) on X for a summary of the system prompt changes in each release.**
@@ -220,10 +220,10 @@ The content of various template files embedded in Claude Code.
- [Data: Cowork plugin examples](./system-prompts/data-cowork-plugin-examples.md) (**3158** tks) - Reference examples of minimal, medium, and complex Cowork plugin structures with plugin metadata, skills, agents, hooks, MCP config, README, and connectors.
- [Data: Cross-session inbound setting](./system-prompts/data-cross-session-inbound-setting.md) (**175** tks) - Describes crossSessionInbound delivery, hold, refusal, and permission-mode parity behavior for peer messages.
- [Data: Data visualization anti-patterns](./system-prompts/data-data-visualization-anti-patterns.md) (**2436** tks) - Reference list of chart and dashboard anti-patterns to check against before shipping a data visualization.
-- [Data: Data visualization choosing a form](./system-prompts/data-data-visualization-choosing-a-form.md) (**1799** tks) - Reference guidance for choosing the appropriate chart, stat tile, or non-chart form based on the data's job.
+- [Data: Data visualization choosing a form](./system-prompts/data-data-visualization-choosing-a-form.md) (**1872** tks) - Reference guidance for choosing the appropriate chart, stat tile, or non-chart form based on the data's job.
- [Data: Data visualization components](./system-prompts/data-data-visualization-components.md) (**899** tks) - Reference specification for the HTML/SVG pieces, layers, and system parameters that make up a data visualization.
- [Data: Data visualization interaction](./system-prompts/data-data-visualization-interaction.md) (**1318** tks) - Reference guidance for chart hover layers, tooltips, filters, linked highlighting, and interaction behavior.
-- [Data: Data visualization marks and anatomy](./system-prompts/data-data-visualization-marks-and-anatomy.md) (**2573** tks) - Reference specifications for chart marks, spacing, labels, axes, legends, and stat-tile anatomy.
+- [Data: Data visualization marks and anatomy](./system-prompts/data-data-visualization-marks-and-anatomy.md) (**2604** tks) - Reference specifications for chart marks, spacing, labels, axes, legends, and stat-tile anatomy.
- [Data: Dialog expiry setting](./system-prompts/data-dialog-expiry-setting.md) (**204** tks) - Describes dialogExpiry deadlines for remote permission dialogs and held cross-session messages, including defaults and overrides.
- [Data: DirectoryAdded hook description](./system-prompts/data-directoryadded-hook-description.md) (**216** tks) - Describes when the DirectoryAdded hook fires, its input fields, and how failures and output are handled for add-dir and register_repo_root sources.
- [Data: Files API reference — C#](./system-prompts/data-files-api-reference-c.md) (**378** tks) - C# Files API reference including beta file upload and use with beta message types.
@@ -251,9 +251,9 @@ The content of various template files embedded in Claude Code.
- [Data: MCP HTTP client capability validation rationale](./system-prompts/data-mcp-http-client-capability-validation-rationale.md) (**189** tks) - Explains where required client capabilities are validated in the modern MCP HTTP request ladder and how the pre-dispatch gate affects observable precedence.
- [Data: MCP HTTP parameter header validation rationale](./system-prompts/data-mcp-http-parameter-header-validation-rationale.md) (**247** tks) - Explains validation of Mcp-Param headers against tool declarations and request arguments, including error shape and pre-dispatch precedence.
- [Data: MCP HTTP standard header validation rationale](./system-prompts/data-mcp-http-standard-header-validation-rationale.md) (**326** tks) - Explains the modern MCP HTTP validation ladder's Mcp-Method and Mcp-Name header checks and their observed pre-dispatch precedence.
-- [Data: MCP call check_permissions field](./system-prompts/data-mcp-call-check-permissions-field.md) (**346** tks) - Documents the mcp_call check_permissions flag that applies normal permission checks and hooks to host-originated calls, its capability and transport restrictions, and its model-context exclusion.
+- [Data: MCP call check_permissions field](./system-prompts/data-mcp-call-check-permissions-field.md) (**376** tks) - Documents the mcp_call check_permissions flag that applies normal permission checks and hooks to host-originated calls, its capability and transport restrictions, and its model-context exclusion.
- [Data: MCP read resource control request](./system-prompts/data-mcp-read-resource-control-request.md) (**437** tks) - Documents the mcp_read_resource control request that reads one MCP Apps (SEP-1865) ui:// resource from a connected MCP server, its sandboxed-render requirement, SDK-server rejection, refused lanes and error cases.
-- [Data: MCP server status error_code field](./system-prompts/data-mcp-server-status-error-code-field.md) (**365** tks) - Describes host-actionable MCP connection failure codes, including rejected credentials, required approval and recently unavailable OAuth client registration, with reconnect and sign-in recovery guidance.
+- [Data: MCP server status error_code field](./system-prompts/data-mcp-server-status-error-code-field.md) (**441** tks) - Describes host-actionable MCP connection failure codes, including rejected credentials, required approval and recently unavailable OAuth client registration, with reconnect and sign-in recovery guidance.
- [Data: MCP server status urlHost field](./system-prompts/data-mcp-server-status-urlhost-field.md) (**193** tks) - Documents the settings-file host and port exposed for an HTTP or SSE MCP server when its configuration is withheld, including cloud-hosting and privacy restrictions.
- [Data: MCP status control request liveReply field](./system-prompts/data-mcp-status-control-request-livereply-field.md) (**328** tks) - Describes the mcp_status control request liveReply flag, which publishes the success reply only to currently attached clients through CCR without storing it, and the loss and size-limit caveats.
- [Data: Managed Agents quickstart — Bug hunter](./system-prompts/data-managed-agents-quickstart-bug-hunter.md) (**2585** tks) - Bundled Managed Agents quickstart template for a bug hunter agent that splits a codebase into slices, has parallel readers and second-look reviewers find suspicions, and reports only bugs a separate prover agent demonstrated, with its agent.md definition, environment.yaml and session.yaml.
@@ -292,10 +292,10 @@ The content of various template files embedded in Claude Code.
- [Data: REPL synchronous context API declarations](./system-prompts/data-repl-synchronous-context-api-declarations.md) (**752** tks) - Defines synchronous REPL context APIs and the task notifications delivered when a context becomes idle, fails or is released.
- [Data: Rate limit grace signal](./system-prompts/data-rate-limit-grace-signal.md) (**306** tks) - Describes the usage-limit grace-window signal tracked from the latest served response, including when it sets, clears, and how it interacts with overage status at hard exhaustion.
- [Data: Rate limit unified windows](./system-prompts/data-rate-limit-unified-windows.md) (**453** tks) - Describes per-window subscription rate-limit utilization and reset data derived from unified rate-limit response headers.
-- [Data: Remote tools reannounce control request](./system-prompts/data-remote-tools-reannounce-control-request.md) (**220** tks) - Documents the remote_tools_reannounce control request a replacement worker sends so the attached client re-announces a machine missing from its tool roster, including worker-epoch, once-per-epoch, ten-second and deadline rules.
+- [Data: Remote tools reannounce control request](./system-prompts/data-remote-tools-reannounce-control-request.md) (**353** tks) - Documents the remote_tools_reannounce control request a replacement worker sends so the attached client re-announces a machine missing from its tool roster, including worker-epoch, once-per-epoch, ten-second and deadline rules.
- [Data: Review upload excluded changes error](./system-prompts/data-review-upload-excluded-changes-error.md) (**327** tks) - Error shown when a review upload has no new commits or uploadable changes because every uncommitted file was excluded from the git bundle.
- [Data: Rewind files skippedLinks field](./system-prompts/data-rewind-files-skippedlinks-field.md) (**223** tks) - Describes the rewindFiles skippedLinks count, including link-safety refusal semantics and dry-run behavior.
-- [Data: SDK API error kind field](./system-prompts/data-sdk-api-error-kind-field.md) (**2517** tks) - Documents the typed api_error kinds carried on SDK assistant messages that wrap an API error, and what each kind means for replaying, rewinding, or remedying the request.
+- [Data: SDK API error kind field](./system-prompts/data-sdk-api-error-kind-field.md) (**2510** tks) - Documents the typed api_error kinds carried on SDK assistant messages that wrap an API error, and what each kind means for replaying, rewinding, or remedying the request.
- [Data: SDK MCP server connectorId field](./system-prompts/data-sdk-mcp-server-connectorid-field.md) (**225** tks) - Documents the connector identity declared for an SDK MCP server, its server-capability confirmation, and its restriction-only aliases for permission rules and hooks.
- [Data: SDK MCP server dataFlowServer field](./system-prompts/data-sdk-mcp-server-dataflowserver-field.md) (**192** tks) - Describes the internal MCP server field that tells Claude Code where the host got a server (Anthropic's registry, a claude.ai connector, or user configuration) for the data flow records of calls to it, and when it is read.
- [Data: SDK MCP server disableAutoBackground field](./system-prompts/data-sdk-mcp-server-disableautobackground-field.md) (**180** tks) - Schema description for the SDK MCP server disableAutoBackground option that stops calls to a server's tools from being moved to a background task, for servers whose calls wait on a person.
@@ -303,6 +303,7 @@ The content of various template files embedded in Claude Code.
- [Data: SDK UI attach draws field](./system-prompts/data-sdk-ui-attach-draws-field.md) (**350** tks) - Documents per-client Backdrop and mobile Input and Select rendering declarations, including repeated attachments, unknown names, limits and empty-list behavior.
- [Data: SDK UI prompt paint request schema](./system-prompts/data-sdk-ui-prompt-paint-request-schema.md) (**293** tks) - Documents agent-originated composer paint requests, their response and timeout behavior, and remote-session authorization.
- [Data: SDK api_error_params field](./system-prompts/data-sdk-api-error-params-field.md) (**1087** tks) - Schema description for the api_error_params object, listing the per-kind parameters (effort, media, media_reason, provider, remedy, rate_limit_info) and their meanings.
+- [Data: SDK api_error_type field](./system-prompts/data-sdk-api-error-type-field.md) (**380** tks) - Documents the API response error type exposed to SDK hosts, its omission rules and why it alone does not determine whether a retry is appropriate.
- [Data: SDK can_use_tool decision_reason_code field](./system-prompts/data-sdk-can-use-tool-decision-reason-code-field.md) (**239** tks) - Documents host-actionable reason codes on can_use_tool permission requests, including the MCP organization ask ceiling and exclusion of denial-only auto-mode codes.
- [Data: SDK can_use_tool plugin field](./system-prompts/data-sdk-can-use-tool-plugin-field.md) (**210** tks) - Identifies the engine-supplied plugin name shown for a permission request and explains missing names and safe display.
- [Data: SDK can_use_tool server_policy_ask_only field](./system-prompts/data-sdk-can-use-tool-server-policy-ask-only-field.md) (**457** tks) - Describes the internal can_use_tool field saying whether nothing asked for a person on a call except the default prompt and at most a host-pushed always-ask tool policy, which asks carry it, and what it does not tell the host.
@@ -321,6 +322,7 @@ The content of various template files embedded in Claude Code.
- [Data: SDK interrupt spare_background_tasks field](./system-prompts/data-sdk-interrupt-spare-background-tasks-field.md) (**160** tks) - Documents the per-request interrupt flag that ends the current turn while leaving background agents and workflows running, with observer and server exceptions.
- [Data: SDK message_stop abandoned_blocks field](./system-prompts/data-sdk-message-stop-abandoned-blocks-field.md) (**386** tks) - Schema description for the abandoned_blocks marker on a Claude Code-produced message_stop, telling consumers which streamed content blocks will never get an assistant message and may be removed.
- [Data: SDK model fallback api_refusal_cyber_verification field](./system-prompts/data-sdk-model-fallback-api-refusal-cyber-verification-field.md) (**225** tks) - Documents the API's Cyber Verification Program status on session-level cyber-refusal fallback notices, including multi-hop attribution and unknown-status handling.
+- [Data: SDK oauth_token_refresh refused_call field](./system-prompts/data-sdk-oauth-token-refresh-refused-call-field.md) (**374** tks) - Explains how refused_call distinguishes model, connector and other OAuth failures and whether a host must renew the token or may validate and reuse it.
- [Data: SDK permission_check_status event schema](./system-prompts/data-sdk-permission-check-status-event-schema.md) (**279** tks) - Schema description for the permission_check_status event that tells hosts a tool call is still waiting on the auto-mode classifier, with checking/done pairing and best-effort display-only semantics.
- [Data: SDK plugin_install system message](./system-prompts/data-sdk-plugin-install-system-message.md) (**189** tks) - Describes the headless plugin installation progress message, its started, installed, failed and completed statuses, how synchronous and background installs differ in bracketing and delivery, and that its error text is untrusted.
- [Data: SDK read_file anchor as_written kind](./system-prompts/data-sdk-read-file-anchor-as-written-kind.md) (**567** tks) - Schema description for the read_file anchor's as_written kind, which resolves a path copied from Claude's replies against the cwd, git root, working directories, and previously read or written files, and treats ambiguous, out-of-scope, or unresolved paths as final refusals.
@@ -330,6 +332,7 @@ The content of various template files embedded in Claude Code.
- [Data: SDK read_file lossy field](./system-prompts/data-sdk-read-file-lossy-field.md) (**210** tks) - Documents whether UTF-8 file-read output replaced invalid bytes, distinguishes literal replacement characters from lossy decoding, and explains base64 and older-client absence.
- [Data: SDK read_file offset field](./system-prompts/data-sdk-read-file-offset-field.md) (**319** tks) - Describes the internal read_file request field for the byte a read starts at, counted from the file's start or back from its end, including clamping, how to read on with the response's offset and read_bytes, refused values, and responders that ignore it.
- [Data: SDK read_file requested_path field](./system-prompts/data-sdk-read-file-requested-path-field.md) (**231** tks) - Describes the internal read_file reply field echoing the request's path exactly as sent, so a client can drop a reply that answers another request under the same request_id, and when the field is absent.
+- [Data: SDK reload_plugins refuse_if_busy field](./system-prompts/data-sdk-reload-plugins-refuse-if-busy-field.md) (**293** tks) - Documents the reload_plugins option that rejects a reload while the session has active work and the capability hosts use to detect support.
- [Data: SDK result credential_expires_at field](./system-prompts/data-sdk-result-credential-expires-at-field.md) (**213** tks) - Documents the stored claude.ai login expiration timestamp on numbered results, including refresh-chain lifetime, estimate uncertainty and omission rules.
- [Data: SDK result main stream chunk count field](./system-prompts/data-sdk-result-main-stream-chunk-count-field.md) (**214** tks) - Documents the count of runtime fetch chunks across main streams and when the count is absent.
- [Data: SDK result message_normalization_detail field](./system-prompts/data-sdk-result-message-normalization-detail-field.md) (**523** tks) - Documents the message-normalization timing breakdown, image checks, post-processing and source-versus-wire message counts before a turn's first dispatch.
@@ -349,8 +352,8 @@ The content of various template files embedded in Claude Code.
- [Data: SDK set model success payload model field](./system-prompts/data-sdk-set-model-success-payload-model-field.md) (**235** tks) - Schema description for the internal model field on the set_model success payload, which reports the model the next turn runs spelled as in system/init, including alias resolution, plan-mode and refusal-fallback caveats, and which hosts send it.
- [Data: SDK set model system prompt field](./system-prompts/data-sdk-set-model-system-prompt-field.md) (**245** tks) - Schema description for the internal set_model system_prompt field, including next-turn application with unchanged tool definitions, non-empty updates, feature overrides, and compatibility behavior.
- [Data: SDK set_has_subscribers control request](./system-prompts/data-sdk-set-has-subscribers-control-request.md) (**326** tks) - Documents the set_has_subscribers control request that adjusts stream-json output batching when nobody is watching without dropping events.
-- [Data: SDK startup failure reason codes](./system-prompts/data-sdk-startup-failure-reason-codes.md) (**1240** tks) - Defines known startup refusal reasons so hosts can distinguish retryable failures from sign-in, managed-policy, environment, worktree and initialization problems requiring a fix.
-- [Data: SDK stream event owed_rows field](./system-prompts/data-sdk-stream-event-owed-rows-field.md) (**254** tks) - Describes the internal stream-event field listing the uuids of project_doc_reads rows a turn wrote before its first model request that give the text of an added claude.ai Project document, including what counts as added and when it is present.
+- [Data: SDK startup failure reason codes](./system-prompts/data-sdk-startup-failure-reason-codes.md) (**1256** tks) - Defines known startup refusal reasons so hosts can distinguish retryable failures from sign-in, managed-policy, environment, worktree and initialization problems requiring a fix.
+- [Data: SDK stream event owed_rows field](./system-prompts/data-sdk-stream-event-owed-rows-field.md) (**230** tks) - Documents the stream-event field listing UUIDs of Project document-read and changed Project-context rows written before the first model request, present alongside user_message_uuid.
- [Data: SDK system init plugin_errors field](./system-prompts/data-sdk-system-init-plugin-errors-field.md) (**480** tks) - Schema description for the plugin_errors array on the SDK system/init message, covering failed versus partially loaded plugins, name@marketplace and positional inline/synced tags, open-set error types, omission when clean, and Remote Control workers always omitting it.
- [Data: SDK system init ultrathink_typed_raises field](./system-prompts/data-sdk-system-init-ultrathink-typed-raises-field.md) (**210** tks) - Documents when clients may highlight typed ultrathink and assert ultrathink_typed, including effort-change invalidation and Remote Control bridge limitations.
- [Data: SDK task notification handback field](./system-prompts/data-sdk-task-notification-handback-field.md) (**205** tks) - Schema description for the handback field on a completed task notification from a subagent that reports through the SubagentHandback tool, including when it is omitted and how summary then reads.
@@ -359,11 +362,12 @@ The content of various template files embedded in Claude Code.
- [Data: SDK ui_invalidate event](./system-prompts/data-sdk-ui-invalidate-event.md) (**185** tks) - Documents plugin UI invalidation events, redraw scope, evaluation deduplication and per-plugin notification coalescing.
- [Data: SDK ui_invalidate instances field](./system-prompts/data-sdk-ui-invalidate-instances-field.md) (**244** tks) - Documents the optional list of stale plugin UI instances that clients may redraw independently after state, fault or backdrop changes, with full redraw as the compatible fallback.
- [Data: SDK ui_prompt_autocomplete request schema](./system-prompts/data-sdk-ui-prompt-autocomplete-request-schema.md) (**195** tks) - Schema description for the ui_prompt_autocomplete request in which a remote surface's Composer asks the engine to raise prompt.autocomplete for the token at the caret, including per-client ordering and superseded handling.
-- [Data: SDK ui_render component field](./system-prompts/data-sdk-ui-render-component-field.md) (**215** tks) - Describes the internal ui_render field naming what to draw, including the Theme component that the engine answers itself with the standing colors, color scheme override and still backdrop, and the Backdrop component drawn by a mod's hook.
+- [Data: SDK ui_render component field](./system-prompts/data-sdk-ui-render-component-field.md) (**259** tks) - Describes the internal ui_render field naming what to draw, including the Theme component that the engine answers itself with the standing colors, color scheme override and still backdrop, and the Backdrop component drawn by a mod's hook.
- [Data: SDK user message requeue_if_interrupted_within_ms field](./system-prompts/data-sdk-user-message-requeue-if-interrupted-within-ms-field.md) (**943** tks) - Schema description for the internal user-message requeue_if_interrupted_within_ms input, which puts a prompt back at the front of the queue when an interrupt crosses a turn that was taking it in, including the timing window, interrupt-response reporting, transcript and hook effects, and eligibility rules.
- [Data: SDK user message retracted_message_uuids field](./system-prompts/data-sdk-user-message-retracted-message-uuids-field.md) (**188** tks) - Describes the internal output-only field on the synthetic re-ask message after an unparseable tool call, listing the uuids of the dropped response messages that a client should remove, and when it is absent.
- [Data: SDK user message ultrathink_typed field](./system-prompts/data-sdk-user-message-ultrathink-typed-field.md) (**257** tks) - Documents the sender's claim that ultrathink was typed without pasted or suggested text, the capability required to send it, and the additional checks before raising turn effort.
-- [Data: SDK voice-call message origin schema](./system-prompts/data-sdk-voice-call-message-origin-schema.md) (**361** tks) - Documents the host-supplied voice-call message origin, the synthetic text-only input shape it requires, its background-only framing, and its storage and replay behavior.
+- [Data: SDK voice-call message origin schema](./system-prompts/data-sdk-voice-call-message-origin-schema.md) (**391** tks) - Documents the host-supplied voice-call message origin, the synthetic text-only input shape it requires, its background-only framing, and its storage and replay behavior.
+- [Data: SDK warning hook_no_verdict field](./system-prompts/data-sdk-warning-hook-no-verdict-field.md) (**212** tks) - Documents structured warning details for a PreToolUse hook that gave no verdict without refusing the call, including timeout and plugin fields.
- [Data: SDK warning model_restriction field](./system-prompts/data-sdk-warning-model-restriction-field.md) (**312** tks) - Documents structured requested and effective model names on model-restriction warnings, their sanitization and trust rules, and startup versus later-switch provenance.
- [Data: SSL cross-certificate missing root error](./system-prompts/data-ssl-cross-certificate-missing-root-error.md) (**574** tks) - API connection error for a certificate chain that runs through a cross-certificate whose issuing root is missing or expired, giving the two alternative fixes in self-service or ask-your-IT-team wording and the unrelated-CA case.
- [Data: SSL name constraint unsupported type explanation](./system-prompts/data-ssl-name-constraint-unsupported-type-explanation.md) (**171** tks) - Explains a certificate refusal caused by a certificate authority whose name constraint covers a kind of name the client cannot check, usually IP addresses, and how the server's operator can fix it with a different certificate or chain.
@@ -373,10 +377,14 @@ The content of various template files embedded in Claude Code.
- [Data: Sandbox credential mask empty injectHosts warning](./system-prompts/data-sandbox-credential-mask-empty-injecthosts-warning.md) (**174** tks) - Warns that sandbox credential masks with empty injectHosts expose only sentinel values and explains how to resolve adapter, managed-settings, and local configuration causes.
- [Data: Sandbox deniedResolvedAddresses setting](./system-prompts/data-sandbox-deniedresolvedaddresses-setting.md) (**209** tks) - Schema description for the sandbox deniedResolvedAddresses setting listing IP addresses or CIDR ranges an allowed hostname must not resolve to, its interaction with allowedDomains IP literals, and its exemption for parentProxy and mitmProxy routes.
- [Data: Sandbox excludedCommands setting](./system-prompts/data-sandbox-excludedcommands-setting.md) (**194** tks) - Describes command-pattern sandbox exclusions, requiring every pipeline or chain component to match while keeping complex calls sandboxed and preserving permission checks.
-- [Data: Sandbox filesystem disabled setting](./system-prompts/data-sandbox-filesystem-disabled-setting.md) (**227** tks) - Describes sandbox.filesystem.disabled behavior, platform limits, read-protection effects, and configuration precedence.
+- [Data: Sandbox failIfUnavailable setting](./system-prompts/data-sandbox-failifunavailable-setting.md) (**193** tks) - Describes when sandbox.failIfUnavailable prevents startup and why hosted cloud containers instead start with a warning.
+- [Data: Sandbox filesystem disabled setting](./system-prompts/data-sandbox-filesystem-disabled-setting.md) (**232** tks) - Describes sandbox.filesystem.disabled behavior, platform limits, read-protection effects, and configuration precedence.
+- [Data: Sandbox linked worktree write denial warning](./system-prompts/data-sandbox-linked-worktree-write-denial-warning.md) (**230** tks) - Warns that a worktrees directory could not be inspected without following a possible remote link and lists the paths denied writes to protect linked worktrees.
- [Data: Sandbox network domain spelling warning](./system-prompts/data-sandbox-network-domain-spelling-warning.md) (**278** tks) - Formats the sandbox diagnostic for unreliable network domain spellings and explains conservative enforcement until entries are corrected.
+- [Data: Sandbox project false value ignored under higher-priority policy](./system-prompts/data-sandbox-project-false-value-ignored-under-higher-priority-policy.md) (**98** tks) - Explains when project-level false values for sandbox.enabled and sandbox.failIfUnavailable are ignored because a higher-priority source sets true.
- [Data: Sandbox project settings ignored under higher-priority policy](./system-prompts/data-sandbox-project-settings-ignored-under-higher-priority-policy.md) (**78** tks) - Appends the higher-priority sandbox-policy conditions under which project settings are ignored.
- [Data: Sandbox root without CAP_SETFCAP error](./system-prompts/data-sandbox-root-without-cap-setfcap-error.md) (**140** tks) - Linux sandbox dependency error explaining that running as uid 0 without CAP_SETFCAP makes every sandboxed command fail while writing a uid map on kernels 5.12+ and backports, and telling the user to grant CAP_SETFCAP or run as non-root.
+- [Data: Sandbox unchecked protected paths warning](./system-prompts/data-sandbox-unchecked-protected-paths-warning.md) (**273** tks) - Warns that unchecked protected paths disable project-supplied filesystem grants and directs the user to remove unexpected links or use user settings.
- [Data: Sandbox worktrees directory listing failure warning](./system-prompts/data-sandbox-worktrees-directory-listing-failure-warning.md) (**257** tks) - Warning logged when the sandbox cannot list a repository's worktrees folder, saying whether sandboxed commands are write-denied under it or may change linked worktrees' commondir and config.worktree files until it can be listed again and the sandbox config is rebuilt.
- [Data: Self-hosted runner Anthropic git proxy credential warning](./system-prompts/data-self-hosted-runner-anthropic-git-proxy-credential-warning.md) (**338** tks) - Runner warning that sources on hosts not routed through Anthropic-managed git clone with this machine's credentials, which --use-anthropic-git-proxy wipes from HOME-level git config, so credentials must live in /etc/gitconfig, GIT_ASKPASS, or an SSH key.
- [Data: Self-hosted runner GIT_ASKPASS governed hosts warning](./system-prompts/data-self-hosted-runner-git-askpass-governed-hosts-warning.md) (**311** tks) - Runner warning that a GIT_ASKPASS program answers for every host, so later git config could send this machine's credential for governed hosts through Anthropic's relay, and directs scoping or unsetting it along with core.askPass and SSH_ASKPASS.
@@ -405,13 +413,16 @@ The content of various template files embedded in Claude Code.
- [Data: Structured usage rate-limit rows field](./system-prompts/data-structured-usage-rate-limit-rows-field.md) (**177** tks) - Schema description for structured usage rate-limit rows, including server-defined meter ordering, null and empty semantics, and the synthesized header-fallback row.
- [Data: Submit feedback control request](./system-prompts/data-submit-feedback-control-request.md) (**269** tks) - Documents the internal submit_feedback control request that sends a /feedback report with the session transcript and sanitized error log, returns unavailable_reason when feedback is disabled or unsendable, and honors save_locally by writing a redacted local bundle instead.
- [Data: Telemetry variables ignored notice](./system-prompts/data-telemetry-variables-ignored-notice.md) (**268** tks) - Settings-status warning listing OTEL/telemetry environment variables a settings file sets that Claude Code ignores because such files can only turn telemetry off, with guidance on where to set them intentionally.
-- [Data: Teleport stash index update failure error](./system-prompts/data-teleport-stash-index-update-failure-error.md) (**278** tks) - Teleport error saying the changes were stashed but the checkout's index could not be updated to match, so git status will show phantom changes, with steps to check .git/index, run git reset, and run /teleport again.
+- [Data: Teleport stash MERGE_AUTOSTASH refusal](./system-prompts/data-teleport-stash-merge-autostash-refusal.md) (**199** tks) - Refuses to stash when MERGE_AUTOSTASH exists and explains how to finish an intended merge or remove an unexpected record before retrying teleport.
+- [Data: Teleport stash index update failure error](./system-prompts/data-teleport-stash-index-update-failure-error.md) (**240** tks) - Teleport error saying the changes were stashed but the checkout's index could not be updated to match, so git status will show phantom changes, with steps to check .git/index, run git reset, and run /teleport again.
+- [Data: Teleport stash ref without entries refusal](./system-prompts/data-teleport-stash-ref-without-entries-refusal.md) (**248** tks) - Refuses to stash when refs/stash exists without listed entries and provides inspection and recovery commands before teleport is retried.
+- [Data: Teleport stash unexpected entry refusal](./system-prompts/data-teleport-stash-unexpected-entry-refusal.md) (**484** tks) - Explains how to inspect and recover the top stash when its entries change unexpectedly or cannot be read, with extra repair instructions for unsafe Git files.
- [Data: Teleport stash unsafe object store or ref table refusal](./system-prompts/data-teleport-stash-unsafe-object-store-or-ref-table-refusal.md) (**204** tks) - Teleport error saying nothing was stashed because an object store or ref table entry is a link, unreadable or otherwise not what git keeps there, with guidance not to delete what it leads to and to stash or commit manually before running /teleport again.
- [Data: Teleport stash unsafe ref file refusal](./system-prompts/data-teleport-stash-unsafe-ref-file-refusal.md) (**256** tks) - Teleport error saying nothing was stashed because a ref file or a folder on the way to it is a link, has a second name, names another ref, or is not a plain file, with safe repair steps and a warning never to delete HEAD or plain ref files.
- [Data: Thin-client diff dialog noDiffReason field](./system-prompts/data-thin-client-diff-dialog-nodiffreason-field.md) (**358** tks) - Explains the reasons a workspace diff is absent and the retry behavior for missing, refused, busy, failed or moved repositories.
- [Data: Thin-client diff dialog schema](./system-prompts/data-thin-client-diff-dialog-schema.md) (**176** tks) - Internal data description for workspace git diff payloads used by the thin-client diff dialog.
- [Data: Token counting reference](./system-prompts/data-token-counting-reference.md) (**665** tks) - Reference documentation for counting Claude model tokens with the Messages count_tokens endpoint and Anthropic SDK or CLI examples, including warnings against OpenAI tokenizers.
-- [Data: Tool result display metadata field](./system-prompts/data-tool-result-display-metadata-field.md) (**1247** tks) - Documents wrapper-level display metadata for SDK clients, including non-execution reasons, remedies, permission decisions, and human denial feedback, never replayed to the model.
+- [Data: Tool result display metadata field](./system-prompts/data-tool-result-display-metadata-field.md) (**1471** tks) - Documents wrapper-level display metadata for SDK clients, including non-execution reasons, remedies, permission decisions, and human denial feedback, never replayed to the model.
- [Data: Tool use display metadata field](./system-prompts/data-tool-use-display-metadata-field.md) (**787** tks) - Documents the tool_use_meta wire field carrying per-block display metadata for a message's tool_use blocks; it is wrapper-level UI metadata and is not replayed to the model.
- [Data: Tool use reference — C#](./system-prompts/data-tool-use-reference-c.md) (**2893** tks) - C# tool use reference including defining tools and reconstructing response content for the follow-up assistant message.
- [Data: Tool use reference — Ruby](./system-prompts/data-tool-use-reference-ruby.md) (**943** tks) - Ruby tool use reference including the beta tool runner and manual agentic-loop guidance.
@@ -442,7 +453,7 @@ The content of various template files embedded in Claude Code.
- [Data: Worktree isolation unreadable host registry refusal](./system-prompts/data-worktree-isolation-unreadable-host-registry-refusal.md) (**301** tks) - Error text refusing an isolation worktree whose host registry cannot be read safely and requiring manual inspection before git worktree commands.
- [Data: Worktree isolation unverified host repository refusal](./system-prompts/data-worktree-isolation-unverified-host-repository-refusal.md) (**410** tks) - Error text refusing an isolation worktree when its enclosing checkout cannot be verified, listing broken git metadata causes and ways to repair or avoid that layout.
- [Data: Worktree registry inspection and cleanup guidance](./system-prompts/data-worktree-registry-inspection-and-cleanup-guidance.md) (**348** tks) - Recovery guidance requiring manual removal of planted links and duplicate or stray registry entries before pruning, unlocking, and recreating a worktree.
-- [Data: WrapUp tool result nextStep field](./system-prompts/data-wrapup-tool-result-nextstep-field.md) (**445** tks) - Describes the WrapUp result field that carries the next-step box for a client to draw without reading the tool's input, when to open the box and trust it, and how its recap and question are length-limited, redacted and cleaned.
+- [Data: WrapUp tool result nextStep field](./system-prompts/data-wrapup-tool-result-nextstep-field.md) (**436** tks) - Describes the WrapUp result field that carries the next-step box for a client to draw without reading the tool's input, when to open the box and trust it, and how its recap and question are length-limited, redacted and cleaned.
- [Data: allowManagedPermissionRulesOnly allowed-tools frontmatter scope](./system-prompts/data-allowmanagedpermissionrulesonly-allowed-tools-frontmatter-scope.md) (**402** tks) - Continues the allowManagedPermissionRulesOnly schema description: skill, command, and plugin allowed-tools frontmatter is ignored except for managed and bundled skills and plugins on admin-backed channels such as pinned managed marketplaces, attested claude.ai-synced plugins, and host-delivered --plugin-dir plugins.
- [Data: allowManagedPermissionRulesOnly deny rules still apply](./system-prompts/data-allowmanagedpermissionrulesonly-deny-rules-still-apply.md) (**45** tks) - Closes the allowManagedPermissionRulesOnly schema description by noting that --disallowedTools, skill disallowed-tools, and other command-line or session deny and ask rules still apply.
- [Data: allowManagedPermissionRulesOnly setting](./system-prompts/data-allowmanagedpermissionrulesonly-setting.md) (**66** tks) - Schema description for the managed allowManagedPermissionRulesOnly setting, which ignores permission rules from user, project, local, and --settings files and --allowedTools allow rules so only managed settings can add allow rules.
@@ -454,6 +465,7 @@ The content of various template files embedded in Claude Code.
- [Data: blockWritesOutsideWorkingDirectories setting](./system-prompts/data-blockwritesoutsideworkingdirectories-setting.md) (**321** tks) - Describes the file-tool write boundary, session-file exceptions, settings-file protection and worktree-tool restrictions without claiming to constrain shell or MCP execution.
- [Data: deniedModels setting](./system-prompts/data-deniedmodels-setting.md) (**222** tks) - Schema description for the managed deniedModels setting, covering family-alias blocking, version-wide model ID blocking across dates, -fast and provider prefixes, ignored release-dependent aliases, and Default-model step-down or startup failure.
- [Data: denyInsteadOfPrompting setting](./system-prompts/data-denyinsteadofprompting-setting.md) (**276** tks) - Documents the permissions.denyInsteadOfPrompting setting that immediately denies otherwise interactive tool requests while preserving permission rules and hooks.
+- [Data: disableClaudeAiConnectors setting](./system-prompts/data-disableclaudeaiconnectors-setting.md) (**170** tks) - Describes how disableClaudeAiConnectors blocks cloud-connector connections and how true values across settings sources take precedence.
- [Data: sshHostCommands setting](./system-prompts/data-sshhostcommands-setting.md) (**345** tks) - Describes the Claude Desktop sshHostCommands setting, including command syntax, platform requirements, accepted output formats, and trusted settings sources.
- [Data: syncClaudeAiPlugins setting](./system-prompts/data-syncclaudeaiplugins-setting.md) (**332** tks) - Describes how the syncClaudeAiPlugins setting disables account-synced plugin downloads, availability, and cleanup across user, managed, workspace, and invocation scopes.
- [Data: syncClaudeAiSkills setting](./system-prompts/data-syncclaudeaiskills-setting.md) (**332** tks) - Describes how the syncClaudeAiSkills setting disables account-synced skill downloads, availability, and cleanup across user, managed, workspace, and invocation scopes.
@@ -510,7 +522,7 @@ Parts of the main system prompt.
- [System Prompt: Claude Fable 5.1 model identity](./system-prompts/system-prompt-claude-fable-5-1-model-identity.md) (**198** tks) - Identifies this Claude iteration as Claude Fable 5.1, describes the model tier and relationship to Claude Mythos 5.1, and links to the Fable model information page.
- [System Prompt: Claude Fable 5.2 model identity](./system-prompts/system-prompt-claude-fable-5-2-model-identity.md) (**235** tks) - Identifies this Claude iteration as Claude Fable 5.2, describes the model tier and relationship to Claude Mythos 5.2, and links to the Fable model information page.
- [System Prompt: Claude in Chrome browser automation (non-deferred tools)](./system-prompts/system-prompt-claude-in-chrome-browser-automation-non-deferred-tools.md) (**1059** tks) - Guidelines shared by the deferred and directly available Claude in Chrome setups: record multi-step interactions as GIFs, filter console output, avoid blocking JavaScript dialogs, stop and ask instead of looping on failures, and start each session from fresh tab context.
-- [System Prompt: Claude in Chrome browser automation](./system-prompts/system-prompt-claude-in-chrome-browser-automation.md) (**88** tks) - Heading and introduction shared by the deferred and directly available Claude in Chrome setups, telling Claude it has mcp__claude-in-chrome__* tools for working with web pages in Chrome and to follow the guidelines that come after it.
+- [System Prompt: Claude in Chrome browser automation](./system-prompts/system-prompt-claude-in-chrome-browser-automation.md) (**76** tks) - Heading and introduction shared by the deferred and directly available Claude in Chrome setups, telling Claude it has mcp__claude-in-chrome__* tools for working with web pages in Chrome and to follow the guidelines that come after it.
- [System Prompt: Claude in Chrome browser selection instructions](./system-prompts/system-prompt-claude-in-chrome-browser-selection-instructions.md) (**191** tks) - Instructs the agent to ask the user to choose among multiple connected Chrome browsers before using browser automation tools.
- [System Prompt: Claude in Chrome deferred tool loading](./system-prompts/system-prompt-claude-in-chrome-deferred-tool-loading.md) (**271** tks) - Directs Claude to batch-load deferred core and task-specific Chrome MCP tools in one ToolSearch call.
- [System Prompt: Claude in Chrome task-specific tool loading](./system-prompts/system-prompt-claude-in-chrome-task-specific-tool-loading.md) (**75** tks) - Tells Claude to add task-specific Claude in Chrome tools (console and network readers, form input, GIF recording, page scripting) to the same batched ToolSearch call when the task clearly needs them.
@@ -622,10 +634,11 @@ Parts of the main system prompt.
- [System Prompt: Responsive mode](./system-prompts/system-prompt-responsive-mode.md) (**813** tks) - Directs Claude to answer immediately with a one or two sentence acknowledgement before thinking or calling any tool, and to write plain conversational English free of openers, flattery, filler and wrap-ups.
- [System Prompt: Sandbox task resource boundary](./system-prompts/system-prompt-sandbox-task-resource-boundary.md) (**217** tks) - Limits sandboxed work to task-provided resources even when other resources are reachable and requires reporting missing access instead of bypassing the boundary.
- [System Prompt: Saving skills via file delivery](./system-prompts/system-prompt-saving-skills-via-file-delivery.md) (**270** tks) - Explains that account skills are not modified by filesystem writes and directs the agent to deliver complete skill files with the file-sending tool.
+- [System Prompt: Saving skills via review proposals](./system-prompts/system-prompt-saving-skills-via-review-proposals.md) (**233** tks) - Directs skill creation and requested updates through a user-review proposal card, requiring existing skills to be read first and updates to be supplied as edits.
- [System Prompt: Self-hosted runner doctor](./system-prompts/system-prompt-self-hosted-runner-doctor.md) (**8910** tks) - Instructs an agent to diagnose self-hosted runner authentication, network, lifecycle, queue, hooks, metrics, and escalation issues.
- [System Prompt: Self-hosted runner setup](./system-prompts/system-prompt-self-hosted-runner-setup.md) (**2718** tks) - Guides an operator through creating, verifying, learning, and documenting a self-hosted runner deployment.
- [System Prompt: Send-to-project coordinator handoff](./system-prompts/system-prompt-send-to-project-coordinator-handoff.md) (**1141** tks) - Directs the project coordinator to post the transferred brief as a setup card without starting threads or following instructions from the source session.
-- [System Prompt: Session scratchpad temporary-file guidance](./system-prompts/system-prompt-session-scratchpad-temporary-file-guidance.md) (**208** tks) - Requires temporary files to use the session scratchpad unless the user requests otherwise and requires complete absolute paths when referring to its files.
+- [System Prompt: Session scratchpad temporary-file guidance](./system-prompts/system-prompt-session-scratchpad-temporary-file-guidance.md) (**246** tks) - Requires temporary files to use the session scratchpad unless the user requests otherwise and requires complete absolute paths when referring to its files.
- [System Prompt: Shared git stash safety](./system-prompts/system-prompt-shared-git-stash-safety.md) (**201** tks) - Warns that git stash is shared across worktrees and sessions, preferring WIP commits or uniquely tagged stash entries.
- [System Prompt: Skill save permission note](./system-prompts/system-prompt-skill-save-permission-note.md) (**163** tks) - Notes that whether a delivered file can be saved as a skill depends on organization settings the assistant cannot see, and how to phrase that conditionally rather than as a certainty.
- [System Prompt: Skillify Current Session](./system-prompts/system-prompt-skillify-current-session.md) (**1798** tks) - System prompt for converting the current session into a skill.
@@ -662,8 +675,8 @@ Parts of the main system prompt.
- [System Prompt: Windows sandbox localhost workarounds (local binding off)](./system-prompts/system-prompt-windows-sandbox-localhost-workarounds-local-binding-off.md) (**361** tks) - Lists what still works while the Windows sandbox blocks localhost (servers reachable from outside, proxy-honouring clients reaching allowed host:port entries, with the port always kept) and explains that turning on allowLocalBinding, re-running sandbox set-up and restarting Claude Code is the user's decision.
- [System Prompt: Windows sandbox package cache limits](./system-prompts/system-prompt-windows-sandbox-package-cache-limits.md) (**285** tks) - Explains read-only Maven cache access and warns against dotnet no-restore builds that can remove package DLLs.
- [System Prompt: Windows sandbox placeholder file protection](./system-prompts/system-prompt-windows-sandbox-placeholder-file-protection.md) (**187** tks) - Protects sandbox-created zero-byte configuration placeholders from staging, deletion, relocation, and Git ignore changes.
-- [System Prompt: Windows sandbox profile tool retry guidance](./system-prompts/system-prompt-windows-sandbox-profile-tool-retry-guidance.md) (**558** tks) - Identifies profile-installed tool failures as sandbox limits and explains narrow user-controlled read grants alongside retry guidance.
-- [System Prompt: Windows sandbox profile tool user guidance](./system-prompts/system-prompt-windows-sandbox-profile-tool-user-guidance.md) (**536** tks) - Explains narrow user-controlled read grants for profile-installed tools without authorizing sandbox settings changes.
+- [System Prompt: Windows sandbox profile tool retry guidance](./system-prompts/system-prompt-windows-sandbox-profile-tool-retry-guidance.md) (**721** tks) - Identifies profile-installed tool failures as sandbox limits and explains narrow user-controlled read grants alongside retry guidance.
+- [System Prompt: Windows sandbox profile tool user guidance](./system-prompts/system-prompt-windows-sandbox-profile-tool-user-guidance.md) (**699** tks) - Explains narrow user-controlled read grants for profile-installed tools without authorizing sandbox settings changes.
- [System Prompt: Windows sandbox pytest cache hang](./system-prompts/system-prompt-windows-sandbox-pytest-cache-hang.md) (**195** tks) - Explains that pytest can hang after printing its results in the Windows sandbox when it cannot create a private cache folder, and to kill it and rerun it with -p no:cacheprovider.
- [System Prompt: Windows sandbox repository creation prohibition](./system-prompts/system-prompt-windows-sandbox-repository-creation-prohibition.md) (**669** tks) - Forbids creating user-facing repositories in the Windows sandbox because the sandbox account would own them and the user's git would refuse them, and lays out the ordered steps the user should take instead.
- [System Prompt: Windows sandbox server reachability (local binding on)](./system-prompts/system-prompt-windows-sandbox-server-reachability-local-binding-on.md) (**31** tks) - Notes that, with allowLocalBinding on, a server started in the Windows sandbox is also reachable from outside it, such as from the user's browser.
@@ -791,7 +804,7 @@ Text for large system reminders.
- [System Reminder: Lines selected in IDE](./system-prompts/system-reminder-lines-selected-in-ide.md) (**225** tks) - Notification about lines selected by user in IDE.
- [System Reminder: Linked computer plugin server copy guidance](./system-prompts/system-reminder-linked-computer-plugin-server-copy-guidance.md) (**479** tks) - Tells a cloud session linked to the person's computer that a plugin server also runs there, that calls go to the computer's copy by default, when to use the container's copy instead, and to say in a sentence where the work runs.
- [System Reminder: MCP connector call outside session refusal](./system-prompts/system-reminder-mcp-connector-call-outside-session-refusal.md) (**197** tks) - Refuses a connector call made outside the session when its file-path argument record cannot be reached and tells Claude to identify the host program rather than the connector as the fault.
-- [System Reminder: MCP connector file path refusal](./system-prompts/system-reminder-mcp-connector-file-path-refusal.md) (**536** tks) - Refuses a connector file upload when its path is ineligible, explaining read permissions, session-only allowed folders, links, clipboard images, and hosted-session restrictions.
+- [System Reminder: MCP connector file path refusal](./system-prompts/system-reminder-mcp-connector-file-path-refusal.md) (**342** tks) - Refuses a connector file send without revealing whether the path exists and explains session-folder access, read restrictions and how to retry.
- [System Reminder: MCP connector missing conversation messages refusal](./system-prompts/system-reminder-mcp-connector-missing-conversation-messages-refusal.md) (**238** tks) - Refuses unlisted arguments and text outside recognized file-upload arguments when conversation messages are missing and identifies a Claude Code or host fault that a new conversation will not fix.
- [System Reminder: MCP connector name too long for file approval](./system-prompts/system-reminder-mcp-connector-name-too-long-for-file-approval.md) (**488** tks) - Refuses file uploads when the connector's escaped name cannot fit in the approval prompt and tells the user how much to shorten it while leaving file-free calls available.
- [System Reminder: MCP connector unreadable saved tool definitions refusal](./system-prompts/system-reminder-mcp-connector-unreadable-saved-tool-definitions-refusal.md) (**204** tks) - Refuses unlisted arguments and text outside recognized file-upload arguments when saved tool definitions are unreadable and recommends a new conversation.
@@ -816,6 +829,7 @@ Text for large system reminders.
- [System Reminder: Memory file contents (unverified managed folder)](./system-prompts/system-reminder-memory-file-contents-unverified-managed-folder.md) (**83** tks) - Heads an unverified instruction file from this machine's managed folder in the combined memory block, showing its escaped path and a parenthetical warning that this session's own account could have written it, so Claude must follow every limit it sets but treat nothing in it as permission.
- [System Reminder: Memory file contents](./system-prompts/system-reminder-memory-file-contents.md) (**64** tks) - Heads a memory or instruction file's contents in the combined memory block with a parenthetical naming its type — project instructions, private project instructions, auto-memory, organization-managed policy, or private global instructions.
- [System Reminder: Memory index capacity warning](./system-prompts/system-reminder-memory-index-capacity-warning.md) (**330** tks) - Warns when a private or team memory index approaches or exceeds its byte or line read limit and instructs Claude to compact it below the target size.
+- [System Reminder: Memory listing additions](./system-prompts/system-reminder-memory-listing-additions.md) (**190** tks) - Adds memory files missing from the latest listing and requires reading them before editing or creating files at their paths while treating their listing text as data.
- [System Reminder: Memory sync delete not sent](./system-prompts/system-reminder-memory-sync-delete-not-sent.md) (**181** tks) - Tells Claude that memory sync is not sending changes, so a memory delete has probably not gone out and shared memory may still hold the content, and not to tell the user the content is gone or promise a place to delete it.
- [System Reminder: Memory updates](./system-prompts/system-reminder-memory-updates.md) (**237** tks) - Introduces refreshed memory-file state, invalidates earlier reads, and forbids reproducing the machine-generated update block to the user.
- [System Reminder: Nested instruction file contents](./system-prompts/system-reminder-nested-instruction-file-contents.md) (**45** tks) - Frames a nested CLAUDE.md or AGENTS.md instruction file's contents by path when the instruction-file module attaches one found below the session's working directory.
@@ -830,7 +844,7 @@ Text for large system reminders.
- [System Reminder: Output token limit continuation](./system-prompts/system-reminder-output-token-limit-continuation.md) (**57** tks) - Tells Claude its response was cut off by the output token limit and to resume at the start of the line, table row, list item or sentence that was cut, reopening any code fence or table first.
- [System Reminder: Output token limit tool call cutoff](./system-prompts/system-reminder-output-token-limit-tool-call-cutoff.md) (**98** tks) - Tells Claude that its response hit the output token limit inside a tool call, so the call did not run and nothing of it was kept, and to make the call again in a shorter response by splitting long content across smaller calls.
- [System Reminder: Oversized mentioned file reading guidance](./system-prompts/system-reminder-oversized-mentioned-file-reading-guidance.md) (**380** tks) - Directs reading an unattached oversized mentioned file in portions, estimates the required calls and requires disclosure when an answer uses only searches or partial content.
-- [System Reminder: Plan approved](./system-prompts/system-reminder-plan-approved.md) (**154** tks) - Notifies Claude that the user approved the plan, provides the saved plan file path when there is one and the approved plan content, and allows coding to begin.
+- [System Reminder: Plan approved](./system-prompts/system-reminder-plan-approved.md) (**198** tks) - Notifies Claude that the user approved the plan, provides the saved plan file path when there is one and the approved plan content, and allows coding to begin.
- [System Reminder: Plan awaiting team-lead approval](./system-prompts/system-reminder-plan-awaiting-team-lead-approval.md) (**116** tks) - Reminder laying out what happens after a plan is submitted for team-lead approval.
- [System Reminder: Plan file reference](./system-prompts/system-reminder-plan-file-reference.md) (**133** tks) - Reference to an existing plan file.
- [System Reminder: Plan mode approval tool enforcement](./system-prompts/system-reminder-plan-mode-approval-tool-enforcement.md) (**360** tks) - Requires plan mode turns to end with either AskUserQuestion for clarification or ExitPlanMode for plan approval, and forbids asking for approval any other way.
@@ -840,6 +854,7 @@ Text for large system reminders.
- [System Reminder: Plan mode is active](./system-prompts/system-reminder-plan-mode-is-active.md) (**147** tks) - Reminds Claude that plan mode is active, clarifications should use AskUserQuestion, plans should use ExitPlanMode, and edits are not allowed.
- [System Reminder: Plan mode phase 2 design](./system-prompts/system-reminder-plan-mode-phase-2-design.md) (**495** tks) - Plan-mode phase 2 guidance for launching Plan agents to design an implementation approach after initial exploration.
- [System Reminder: Plan mode re-entry](./system-prompts/system-reminder-plan-mode-re-entry.md) (**339** tks) - System reminder sent when the user enters Plan mode after having previously exited it either via shift+tab or by approving Claude's plan.
+- [System Reminder: Post-compaction conversation reread guidance](./system-prompts/system-reminder-post-compaction-conversation-reread-guidance.md) (**207** tks) - Requires rereading earlier conversation text before quoting it after compaction and explains backward pagination and text that may not yet be readable.
- [System Reminder: PreToolUse hook timeout recovery](./system-prompts/system-reminder-pretooluse-hook-timeout-recovery.md) (**153** tks) - Explains hook timeout and missing-decision refusals, recommends one isolated or smaller retry and leaves hook configuration changes to the user.
- [System Reminder: Prefetched Artifact type instruction files](./system-prompts/system-reminder-prefetched-artifact-type-instruction-files.md) (**294** tks) - Reports that an opened Artifact's type instruction files were prefetched to disk and directs when to read the Artifact URL or local instruction pages.
- [System Reminder: Prefetched design system file reading guidance](./system-prompts/system-reminder-prefetched-design-system-file-reading-guidance.md) (**192** tks) - Directs reading prefetched design-system documentation and treats its files as styling data rather than instructions.
@@ -864,7 +879,7 @@ Text for large system reminders.
- [System Reminder: Remote machine branch transfer review](./system-prompts/system-reminder-remote-machine-branch-transfer-review.md) (**1568** tks) - Sets out how work moves between two unsynced project copies through git: push a branch, fetch it on the user's machine, show the full commit id plus two diff commands, flag agent-config and lookalike paths, submodules and checkout-time filters, and leave the decision to the user addressed by commit id only.
- [System Reminder: Remote machine separate project copies](./system-prompts/system-reminder-remote-machine-separate-project-copies.md) (**206** tks) - Tells the session that its own checkout is the primary copy and the folder on the user's machine is a separate unsynced copy that may sit at another commit, hold uncommitted work, or not even be the same repository, and to state which copy any report refers to.
- [System Reminder: Remote machine-only resources routing](./system-prompts/system-reminder-remote-machine-only-resources-routing.md) (**540** tks) - Directs tasks needing platform tools, devices, logins, user paths, large external data, or internal hosts straight to the online attached machine, says where pasted content and Linux-installable tools belong, and rules out routing project build failures or blocked public sites there.
-- [System Reminder: Reserved-address MCP server refusal](./system-prompts/system-reminder-reserved-address-mcp-server-refusal.md) (**472** tks) - Lists MCP servers refused for reserved addresses and explains that retries cannot resolve the refusal.
+- [System Reminder: Reserved-address MCP server refusal](./system-prompts/system-reminder-reserved-address-mcp-server-refusal.md) (**473** tks) - Lists MCP servers refused for reserved addresses and explains that retries cannot resolve the refusal.
- [System Reminder: Resume agents interrupted by container restart](./system-prompts/system-reminder-resume-agents-interrupted-by-container-restart.md) (**195** tks) - Lists interrupted agents whose results were lost and asks the parent to resume them from saved transcripts before relaunching.
- [System Reminder: Resumed command approval directory verification failure](./system-prompts/system-reminder-resumed-command-approval-directory-verification-failure.md) (**194** tks) - Refuses a command approved before a restart when its original directory cannot be verified and directs a fresh call from the intended directory.
- [System Reminder: Sandbox hard-linked guarded file refusal](./system-prompts/system-reminder-sandbox-hard-linked-guarded-file-refusal.md) (**304** tks) - Refuses sandboxed commands and unsandboxed retries while a guarded file has extra hard links, identifies displayable affected locations and explains recovery without restarting.
@@ -877,7 +892,8 @@ Text for large system reminders.
- [System Reminder: Session moved to new runner resume prompt](./system-prompts/system-reminder-session-moved-to-new-runner-resume-prompt.md) (**168** tks) - Resume prompt telling Claude to continue after a session moved to a new runner, verify working directory state, and check what an interrupted tool call did before rerunning it.
- [System Reminder: Session stop hook active](./system-prompts/system-reminder-session-stop-hook-active.md) (**111** tks) - Tells Claude a session-scoped Stop hook condition is active and must be treated as the directive until met.
- [System Reminder: Staged file still copying](./system-prompts/system-reminder-staged-file-still-copying.md) (**218** tks) - Explains that a staged file is still copying and how to wait for it.
-- [System Reminder: Staged files failed to arrive](./system-prompts/system-reminder-staged-files-failed-to-arrive.md) (**329** tks) - Warns that staged file paths may be missing or outdated and adds specific remediation guidance for disk-space and filename-length failures.
+- [System Reminder: Staged files disk-space recovery](./system-prompts/system-reminder-staged-files-disk-space-recovery.md) (**353** tks) - Explains recovery after staged file copies run out of disk space, distinguishing connector downloads that can be fetched again from files that must be reattached or recreated.
+- [System Reminder: Staged files failed to arrive](./system-prompts/system-reminder-staged-files-failed-to-arrive.md) (**306** tks) - Warns that staged file paths may be missing or outdated and adds specific remediation guidance for disk-space and filename-length failures.
- [System Reminder: Staged files still copying](./system-prompts/system-reminder-staged-files-still-copying.md) (**235** tks) - Explains that staged files are still copying and how to wait for them.
- [System Reminder: Still-running tool call](./system-prompts/system-reminder-still-running-tool-call.md) (**239** tks) - Tells Claude that a tool call is still loading while it answers a message the user sent meanwhile, that the result will arrive separately under the same tool_use_id, and that it must not repeat, wait for, or describe the call as cancelled, backgrounded or detached.
- [System Reminder: Stop hook blocking error](./system-prompts/system-reminder-stop-hook-blocking-error.md) (**20** tks) - Error from a blocking hook command.
@@ -885,6 +901,7 @@ Text for large system reminders.
- [System Reminder: Task tools reminder](./system-prompts/system-reminder-task-tools-reminder.md) (**111** tks) - Reminder to use task tracking tools.
- [System Reminder: Team Coordination](./system-prompts/system-reminder-team-coordination.md) (**356** tks) - System reminder for team coordination.
- [System Reminder: Team Shutdown](./system-prompts/system-reminder-team-shutdown.md) (**136** tks) - System reminder for team shutdown.
+- [System Reminder: Teammate spawned metadata](./system-prompts/system-reminder-teammate-spawned-metadata.md) (**234** tks) - Reports internal teammate launch metadata, forbids repeating it to the user and warns when the teammate starts outside the shell's current folder.
- [System Reminder: TodoWrite reminder](./system-prompts/system-reminder-todowrite-reminder.md) (**86** tks) - Reminder to use TodoWrite tool for task tracking.
- [System Reminder: Token usage](./system-prompts/system-reminder-token-usage.md) (**39** tks) - Current token usage statistics.
- [System Reminder: Tool calls lost to session restart](./system-prompts/system-reminder-tool-calls-lost-to-session-restart.md) (**312** tks) - Lists calls from the previous turn that have no result after a session restart and tells Claude to check whether they ran before issuing them again.
@@ -1054,7 +1071,7 @@ Text for large system reminders.
- [Tool Description: SendUserMessage](./system-prompts/tool-description-sendusermessage.md) (**226** tks) - Describes the SendUserMessage tool for sending user-visible Markdown messages and attachments with normal or proactive status.
- [Tool Description: Session artifact watch lifecycle](./system-prompts/tool-description-session-artifact-watch-lifecycle.md) (**821** tks) - Explains live artifact watch setup, merging a newer published version before republishing, optional comment auto-replies, session restoration, status checks, and lifecycle limits.
- [Tool Description: ShowOnboardingRolePicker](./system-prompts/tool-description-showonboardingrolepicker.md) (**38** tks) - ShowOnboardingRolePicker: presents a row of clickable role chips during Cowork onboarding.
-- [Tool Description: Skill proposal rendering](./system-prompts/tool-description-skill-proposal-rendering.md) (**435** tks) - Defines when and how to render complete recurring-procedure skill proposals for user review without writing files, and requires plugin or built-in skills to be proposed as separately named new skills rather than improvements.
+- [Tool Description: Skill proposal rendering](./system-prompts/tool-description-skill-proposal-rendering.md) (**360** tks) - Defines when and how to render complete recurring-procedure skill proposals for user review without writing files, and requires plugin or built-in skills to be proposed as separately named new skills rather than improvements.
- [Tool Description: Skill-scoped git push refusal](./system-prompts/tool-description-skill-scoped-git-push-refusal.md) (**256** tks) - Explains that an active skill rejects dangerous or bypassing git push forms, permits a plain push to the configured remote, and forbids evasion or silent substitution.
- [Tool Description: SubagentHandback](./system-prompts/tool-description-subagenthandback.md) (**215** tks) - Describes the SubagentHandback tool a subagent calls exactly once to deliver its full final report to the agent that spawned it, noting that the call ends the run and plain-text endings are not delivered.
- [Tool Description: SubscribePR](./system-prompts/tool-description-subscribepr.md) (**877** tks) - Describes subscribing the session to a GitHub pull request so CI failures, the all-green check-suite notice, review comments, and close/reopen events arrive as wake messages (or through the session inbox in remote-control project-thread sessions), treating the permission dialog as the confirmation step, and listing the events that are not forwarded and must still be polled.
@@ -1221,7 +1238,7 @@ Built-in skill prompts for specialized tasks.
- [Skill: /doctor /usage flagged check](./system-prompts/skill-doctor-usage-flagged-check.md) (**458** tks) - Adds a /doctor check that reports, row by row with their shares, the issues /usage measured over the last 7 days of local sessions, treating row text as untrusted data, turning rows with a fix into proposed edits (skipping ones an earlier check already covers and leaving files outside the project and ~/.claude to the user), and presenting rows without a fix as advice for the user to run.
- [Skill: /doctor prompt-audit configuration scope](./system-prompts/skill-doctor-prompt-audit-configuration-scope.md) (**532** tks) - Default arguments /doctor prompt-audit hands to the bundled claude-api skill, scoping the prompt audit to Claude Code configuration files that load in this project, excluding settings files and secrets, limiting edits outside the project, and treating audited files as data.
- [Skill: /doctor slash command description](./system-prompts/skill-doctor-slash-command-description.md) (**372** tks) - Trigger description for the /doctor slash command covering setup health, unused extensions, memory cleanup, hooks, updates, and permission prompts.
-- [Skill: /doctor slash command](./system-prompts/skill-doctor-slash-command.md) (**16296** tks) - Diagnostic workflow for auditing and fixing Claude Code installation health, unused context, local memory duplication, hooks, version currency, and permission prompts.
+- [Skill: /doctor slash command](./system-prompts/skill-doctor-slash-command.md) (**16303** tks) - Diagnostic workflow for auditing and fixing Claude Code installation health, unused context, local memory duplication, hooks, version currency, and permission prompts.
- [Skill: /explain-usage measured usage slash command](./system-prompts/skill-explain-usage-measured-usage-slash-command.md) (**998** tks) - Hands Claude the conversation's measured token usage as JSON with group definitions and caveats, asking for one simple chart of effective usage by group and a plain-language explanation.
- [Skill: /explain-usage slash command](./system-prompts/skill-explain-usage-slash-command.md) (**449** tks) - Analyzes the current session transcript into cost-weighted token usage groups, charts the results, and explains them in plain language.
- [Skill: /init CLAUDE.md and skill setup (new version)](./system-prompts/skill-init-claude-md-and-skill-setup-new-version.md) (**7828** tks) - A comprehensive onboarding flow for setting up CLAUDE.md and related skills/hooks in the current repository, including codebase exploration, user interviews, and iterative proposal refinement.
@@ -1239,6 +1256,8 @@ Built-in skill prompts for specialized tasks.
- [Skill: /send-to-project work brief slash command](./system-prompts/skill-send-to-project-work-brief-slash-command.md) (**857** tks) - Loads the context and transfer tools, reads gathered folder context as untrusted data and drafts a work brief for the user's project-transfer confirmation.
- [Skill: /stuck (background-daemon diagnostics)](./system-prompts/skill-stuck-background-daemon-diagnostics.md) (**273** tks) - The background-daemon troubleshooting section of the /stuck skill.
- [Skill: /stuck slash command](./system-prompts/skill-stuck-slash-command.md) (**964** tks) - Diagnoses frozen or slow Claude Code sessions.
+- [Skill: Artifact design direction cards](./system-prompts/skill-artifact-design-direction-cards.md) (**2476** tks) - Lists the pool of typography, colour and layout directions from which three are dealt into the artifact-design skill guide for a new Artifact page.
+- [Skill: Artifact design page contract](./system-prompts/skill-artifact-design-page-contract.md) (**1577** tks) - Defines the Artifact page contract that opens the artifact-design skill: publishing skeleton, allowed script and font sources, sandboxed browser APIs, responsive layout, theming and the single-check authoring workflow.
- [Skill: Artifact diagramming](./system-prompts/skill-artifact-diagramming.md) (**1413** tks) - Provides Artifact diagramming guidance, including when diagrams add value and how to build legible theme-aware inline SVG.
- [Skill: Claude API CLI installation guidance](./system-prompts/skill-claude-api-cli-installation-guidance.md) (**293** tks) - Claude API skill section giving Anthropic CLI installation routes and requiring its full path when Apache Ant shares the ant command name.
- [Skill: Claude API Managed Agents onboarding from URL unavailable](./system-prompts/skill-claude-api-managed-agents-onboarding-from-url-unavailable.md) (**171** tks) - Section appended to the Claude API skill when its reference files could not be written to disk, telling Claude to say that onboarding an agent from a URL is unavailable, that running /claude-api again retries saving the files, and to stop without fetching or writing anything.
@@ -1266,14 +1285,13 @@ Built-in skill prompts for specialized tasks.
- [Skill: Code Review low effort expanded-findings mode](./system-prompts/skill-code-review-low-effort-expanded-findings-mode.md) (**810** tks) - Low-effort /code-review prompt that reads the diff once, returns up to eight hunk-visible findings, and targets at least min(files_changed, 4) genuine findings.
- [Skill: Computer Use MCP](./system-prompts/skill-computer-use-mcp.md) (**1731** tks) - Instructions for using computer-use MCP tools including tool selection tiers, app access tiers, link safety, and financial action restrictions.
- [Skill: Data Visualization description](./system-prompts/skill-data-visualization-description.md) (**496** tks) - Trigger description for the Data Visualization skill covering charts, graphs, dashboards, palettes, stat tiles, and related visualization requests.
-- [Skill: Data Visualization](./system-prompts/skill-data-visualization.md) (**2943** tks) - Skill instructions for producing accessible, brand-neutral charts, graphs, dashboards, and data visualizations using a validated method.
- [Skill: Debugging](./system-prompts/skill-debugging.md) (**653** tks) - Instructions for debugging an issue that the user is encountering in the Claude Code session.
- [Skill: Dynamic pacing loop execution (re-arm decision)](./system-prompts/skill-dynamic-pacing-loop-execution-re-arm-decision.md) (**411** tks) - Re-arm decision step for dynamic-pacing /loop ticks: call ScheduleWakeup with a fallback-heartbeat or observed-cadence delay, a short reason, the dynamic-mode sentinel prompt, and a noop flag, or stop the loop instead.
- [Skill: Dynamic pacing loop execution](./system-prompts/skill-dynamic-pacing-loop-execution.md) (**664** tks) - Step-by-step instructions for executing a dynamic pacing loop that runs tasks, arms and re-arms event monitors, schedules fallback heartbeat ticks, and handles task notifications.
-- [Skill: Generate permission allowlist from transcripts](./system-prompts/skill-generate-permission-allowlist-from-transcripts.md) (**3419** tks) - Analyzes session transcripts to extract frequently used read-only tool-call patterns and adds them to the project's .claude/settings.json permission allowlist to reduce permission prompts.
+- [Skill: Generate permission allowlist from transcripts](./system-prompts/skill-generate-permission-allowlist-from-transcripts.md) (**3449** tks) - Analyzes session transcripts to extract frequently used read-only tool-call patterns and adds them to the project's .claude/settings.json permission allowlist to reduce permission prompts.
- [Skill: Import to Claude Code](./system-prompts/skill-import-to-claude-code.md) (**418** tks) - Generated SKILL.md instructing Claude to finish importing leftover foreign-agent config that `claude import` could not map automatically.
- [Skill: Mod GitHub sharing consent](./system-prompts/skill-mod-github-sharing-consent.md) (**315** tks) - Offers sharing a working mod once and requires separate visibility consent and file inspection before creating a GitHub repository.
-- [Skill: Mod look reference](./system-prompts/skill-mod-look-reference.md) (**2538** tks) - Reference for the mod skill on giving a session a look in the desktop and mobile apps: building theme colors with makeColors, anchors and roles, moving scenes and still backdrops, what good looks like, and what fails without an error.
+- [Skill: Mod look reference](./system-prompts/skill-mod-look-reference.md) (**2936** tks) - Reference for the mod skill on giving a session a look in the desktop and mobile apps: building theme colors with makeColors, anchors and roles, moving scenes and still backdrops, what good looks like, and what fails without an error.
- [Skill: Mod plugin packaging (any plan)](./system-prompts/skill-mod-plugin-packaging-any-plan.md) (**79** tks) - Comes after the mod packaging offer when the plan is unknown or not Team, Enterprise, Pro or Max: after saving, the person presses Share where it appears on the file's card, or sends the file for the recipient to add under Customize or Plugins with Add, then Upload plugin.
- [Skill: Mod plugin packaging (organization plans)](./system-prompts/skill-mod-plugin-packaging-organization-plans.md) (**77** tks) - Comes after the mod packaging offer on Team and Enterprise plans: after saving, the person presses Share on the file's card to add teammates or, where the organization allows, publish to it, and where no Share appears sends the file for the recipient to add with Upload plugin.
- [Skill: Mod plugin packaging (personal plans)](./system-prompts/skill-mod-plugin-packaging-personal-plans.md) (**63** tks) - Comes after the mod packaging offer on Pro and Max plans: after saving, the person sends the file, downloaded from its card or the saved plugin's page, for the recipient to add under Customize or Plugins with Add, then Upload plugin.
diff --git a/system-prompts/data-data-visualization-choosing-a-form.md b/system-prompts/data-data-visualization-choosing-a-form.md
index 29b22b9..32455be 100644
--- a/system-prompts/data-data-visualization-choosing-a-form.md
+++ b/system-prompts/data-data-visualization-choosing-a-form.md
@@ -1,7 +1,7 @@
# Choosing a form
@@ -56,8 +56,8 @@ limit is a **Meter**, above). Answer it with one of these; the donut stays depri
| The reader must... | Use |
|---|---|
-| Compare the parts with each other, rank them, tell close values apart, or read more than 6 parts | **Ranked horizontal bars**, one per part, largest first, one hue: bars on one baseline compare more exactly than segments laid end to end |
-| See how one whole splits, at a glance: one bar alone in a card, no figures listed beside it | **Tall segmented bar**: each part its own block, its share or amount printed on it where it fits, and a legend of names only. <= 6 parts; fold a tail nobody needs to read into "Other" |
+| Compare the parts with each other, rank them, tell close values apart, or read more than 6 parts | **Ranked horizontal bars**, one per part, largest first (parts with a natural order, such as age bands, keep that order), one hue: bars on one baseline compare more exactly than segments laid end to end |
+| See how one whole splits, at a glance: one bar alone in a card, no figures listed beside it | **Tall segmented bar**: each part its own block, its share or amount printed on it where it fits, and a legend of names. <= 6 parts; fold a tail nobody needs to read into "Other" |
| Read the split beside its figures: down the rows of a table or list, or above a list that already gives each part's name and amount | **Thin segmented bar**, no text: the text beside it carries the figures the reader needs, and printing them on the bar adds nothing. One legend for the whole table; a swatch on each list row replaces the legend |
| See the split change over time, or compare several wholes on one shared axis | **Stacked bar chart** (go horizontal for many / long-named categories): its axis carries the values |
@@ -72,5 +72,7 @@ Specs for the two segmented bars -> *Figures* in `marks-and-anatomy.md`.
| 5-6 | soft cap; legend or small multiples |
| 7-8 | token ceiling; past it, fold the tail into "Other," facet into small multiples, or use composite encoding (hue × shape) |
+**Exception to direct labels:** the tall segmented bar takes no name labels; its legend names the parts (*Figures* in `marks-and-anatomy.md`).
+
Never solve "too many series" by generating more hues. A generated 9th hue is
indistinguishable from an existing one under CVD and breaks every check.
diff --git a/system-prompts/data-data-visualization-marks-and-anatomy.md b/system-prompts/data-data-visualization-marks-and-anatomy.md
index f0f6ee6..91f1248 100644
--- a/system-prompts/data-data-visualization-marks-and-anatomy.md
+++ b/system-prompts/data-data-visualization-marks-and-anatomy.md
@@ -1,7 +1,7 @@
# Marks & anatomy
@@ -87,9 +87,10 @@ one swatch restates the title and costs space.
**Thin:** 8px, or 4px in dense rows; fully rounded ends; no text. **Tall:** 28px and never
more; each part its own block, 4px corners, its share (46%) or rounded amount (413M)
printed inside at the leading edge, in white or near-black (*Labels & legend* says which).
- A block too narrow for its label prints none: that value stays in the table view.
+ A block too narrow for its label prints none: that value stays in the table view or, in a
+ static image (no hover, no table view), goes beside the part's name in the legend.
Hover, keyboard focus, and the bar's accessible name give every part's name, amount, and
- share. The legend lists the names in block order and repeats no numbers. The tall bar is
+ share. The legend lists the names in block order and repeats no number a block prints. The tall bar is
exempt from four chart rules: the 24px bar cap, the ban on large saturated blocks, the ban
on a number on every point, and direct name labels (its legend names the parts). It keeps
its legend and its table view.
diff --git a/system-prompts/data-disableclaudeaiconnectors-setting.md b/system-prompts/data-disableclaudeaiconnectors-setting.md
new file mode 100644
index 0000000..e89b727
--- /dev/null
+++ b/system-prompts/data-disableclaudeaiconnectors-setting.md
@@ -0,0 +1,6 @@
+
+When true in any settings source, claude.ai MCP cloud connectors are not auto-fetched or connected, and a claudeai-proxy server passed explicitly (e.g. via --mcp-config or the SDK mcpServers option) does not connect either. Where a managed or served source sets it, the connectors that the desktop app's Code tab hands to a session do not connect either, from the app version that shows them as turned off. Any-source-true wins: a project can opt out, but a project-level false cannot override a user-level true.
diff --git a/system-prompts/data-mcp-call-check-permissions-field.md b/system-prompts/data-mcp-call-check-permissions-field.md
index 1899cb3..70dcfa0 100644
--- a/system-prompts/data-mcp-call-check-permissions-field.md
+++ b/system-prompts/data-mcp-call-check-permissions-field.md
@@ -1,6 +1,6 @@
-The call is someone else's (a page a host renders), so it is decided as the model's own call to this tool is: permission rules under every spelling, the permission mode, requiresUserInteraction, managed settings, and PreToolUse and PostToolUse hooks, which see the arguments and the result. An ask reaches the host as a can_use_tool control request; where no host can answer, and in dontAsk or plan mode, it is refused. A refused call never reaches the server: on a plain call the error subtype, on a staged call staging error_code=tool_error, each with text starting 'The tool did not run: '. The other refusals of mcp_call come first and keep their own words. A PostToolUse hook's rewrite replaces content and drops structuredContent and _meta. The result stays out of the model's context. Only a CLI that says mcp_call_check_permissions_v1 honours this member: an older one ignores it and runs the tool with no check. Honoured only on the pipes of the host that spawned the process; elsewhere the call is refused. A value that is not a boolean is refused.
+The call is someone else's (a page a host renders), so it is decided as the model's own call to this tool is, except that a slash command's allowed-tools and connector card answers never allow it, even during a turn: permission rules under every spelling, the permission mode, requiresUserInteraction, managed settings, and PreToolUse and PostToolUse hooks, which see the arguments and the result. An ask reaches the host as a can_use_tool control request; where no host can answer, and in dontAsk or plan mode, it is refused. A refused call never reaches the server: on a plain call the error subtype, on a staged call staging error_code=tool_error, each with text starting 'The tool did not run: '. The other refusals of mcp_call come first and keep their own words. A PostToolUse hook's rewrite replaces content and drops structuredContent and _meta. The result stays out of the model's context. Only a CLI that says mcp_call_check_permissions_v1 honours this member: an older one ignores it and runs the tool with no check. Honoured only on the pipes of the host that spawned the process; elsewhere the call is refused. A value that is not a boolean is refused.
diff --git a/system-prompts/data-mcp-server-status-error-code-field.md b/system-prompts/data-mcp-server-status-error-code-field.md
index 3fdfff8..427ee02 100644
--- a/system-prompts/data-mcp-server-status-error-code-field.md
+++ b/system-prompts/data-mcp-server-status-error-code-field.md
@@ -1,6 +1,6 @@
-@internal Why a row is not connected, where a host can act on it. On a 'failed' row: CLAUDEAI_BEARER_REJECTED (a claude.ai connector rejected the claude.ai login — sign in again), FIRST_PARTY_AUTH_REJECTED (a first-party Anthropic server rejected its credential — the claude.ai login, or for the Claude Design server its /design-login authorization; the row's error names which sign-in fixes it, except over a transport that stores its frames, where it reads 'Connection failed'), APPROVAL_REQUIRED (a project-scoped server awaiting the user's approval). On a 'failed' or a 'needs-auth' row: CLIENT_REGISTRATION_UNAVAILABLE (the last look at the server's authorization server, under 30 days ago, found no way to register a client, so mcp_authenticate is expected to fail; the row keeps this value until the server is dialed again, for example by mcp_reconnect or a completed sign-in, so keep sign-in and reconnect available). Absent in every other case; new values are additive.
+@internal Why a row is not connected, where a host can act on it. On a 'failed' row: CLAUDEAI_BEARER_REJECTED (a claude.ai connector rejected the claude.ai login — sign in again), FIRST_PARTY_AUTH_REJECTED (a first-party Anthropic server rejected its credential — the claude.ai login, or for the Claude Design server its /design-login authorization; the row's error names which sign-in fixes it, except over a transport that stores its frames, where it reads 'Connection failed'), APPROVAL_REQUIRED (a project-scoped server awaiting the user's approval), CLAUDEAI_CONNECTORS_DISABLED (a claude.ai connector that the desktop app handed over in process, left out because a managed or served source sets disableClaudeAiConnectors: nothing to reconnect). On a 'failed' or a 'needs-auth' row: CLIENT_REGISTRATION_UNAVAILABLE (the last look at the server's authorization server, under 30 days ago, found no way to register a client, so mcp_authenticate is expected to fail; the row keeps this value until the server is dialed again, for example by mcp_reconnect or a completed sign-in, so keep sign-in and reconnect available). Absent in every other case; new values are additive.
diff --git a/system-prompts/data-remote-tools-reannounce-control-request.md b/system-prompts/data-remote-tools-reannounce-control-request.md
index 2d7b385..9e1ef76 100644
--- a/system-prompts/data-remote-tools-reannounce-control-request.md
+++ b/system-prompts/data-remote-tools-reannounce-control-request.md
@@ -1,6 +1,6 @@
-@internal A replacement worker asks the attached client that serves this session's tools to announce it, when a call names a machine its roster lacks. Not addressed to an instance: the worker knows none. A client that serves the machine announces, overriding once what it believes the worker already holds; it honours one ask per epoch, another only when the announce that one made failed or went unanswered, and at most one per ten seconds, and does nothing for an ask it cannot date or read, or whose sender it would refuse a tool call from. Any other client, and a client that predates the request, does nothing or error-replies, and the worker reads no reply: it withdraws the request with control_cancel_request when its wait ends.
+@internal A replacement worker asks the attached client that serves this session's tools to announce it, when a call names a machine its roster lacks. Not addressed to an instance: the worker knows none. A client that serves the machine announces, overriding once what it believes the worker already holds; it honours one ask per epoch, another only when the announce that one made failed or went unanswered, and at most one per ten seconds, and does nothing for an ask it cannot date or read, or whose sender it would refuse a tool call from. Any other client, and a client that predates the request, does nothing or error-replies, and the worker reads no reply: it withdraws the request with control_cancel_request when its wait ends. A worker that the desktop app hosts also asks the app this way, as it starts and at a turn's start while it holds no host, and withdraws none of those asks. The session service answers such an ask with the error device_announce_refused where it would refuse the app's announce now, or device_stream_closed where the device's stream to it is closed; the worker reads either reply to its newest ask alone, and it ends that turn's wait for the announce.
diff --git a/system-prompts/data-sandbox-failifunavailable-setting.md b/system-prompts/data-sandbox-failifunavailable-setting.md
new file mode 100644
index 0000000..8203847
--- /dev/null
+++ b/system-prompts/data-sandbox-failifunavailable-setting.md
@@ -0,0 +1,6 @@
+
+Exit with an error at startup if sandbox.enabled is true but the sandbox cannot start (missing dependencies or unsupported platform). When false (default), a warning is shown and commands run unsandboxed. Intended for managed-settings deployments that require sandboxing as a hard gate. A Claude Code on the web, scheduled-task or Cowork cloud session (Anthropic's cloud container, which has no command sandbox) starts anyway with a warning: the container isolates it, and the sandbox.filesystem and sandbox.network rules in settings are not enforced there.
diff --git a/system-prompts/data-sandbox-filesystem-disabled-setting.md b/system-prompts/data-sandbox-filesystem-disabled-setting.md
index 141dce7..00b7b90 100644
--- a/system-prompts/data-sandbox-filesystem-disabled-setting.md
+++ b/system-prompts/data-sandbox-filesystem-disabled-setting.md
@@ -1,6 +1,6 @@
-Sandboxed commands get unrestricted read/write access to the host filesystem; network egress is still confined to network.allowedDomains. Intended for deployments whose goal is egress control rather than filesystem containment. Does not change Bash prompting: sandbox.autoAllowBashIfSandboxed is independent and still defaults to true, so set it to false to keep prompting for sandboxed commands. Drops the read protection from filesystem.denyRead and credentials.files deny entries for sandboxed commands, since both are enforced by the filesystem layer this turns off; credentials.files mask entries (sentinel binds) and credentials.envVars deny/mask are unaffected.
+Sandboxed commands get unrestricted read/write access to the host filesystem; network egress is still confined to network.allowedDomains. Intended for deployments whose goal is egress control rather than filesystem containment. Does not change Bash prompting: sandbox.autoAllowBashIfSandboxed is independent and still defaults to false, so sandboxed commands keep going through the permission flow unless it is set to true. Drops the read protection from filesystem.denyRead and credentials.files deny entries for sandboxed commands, since both are enforced by the filesystem layer this turns off; credentials.files mask entries (sentinel binds) and credentials.envVars deny/mask are unaffected.
diff --git a/system-prompts/data-sandbox-linked-worktree-write-denial-warning.md b/system-prompts/data-sandbox-linked-worktree-write-denial-warning.md
new file mode 100644
index 0000000..91c415f
--- /dev/null
+++ b/system-prompts/data-sandbox-linked-worktree-write-denial-warning.md
@@ -0,0 +1,10 @@
+
+[Sandbox] ${WORKTREES_DIRECTORY_PATH} is, or is inside, a link that leads or may lead to another computer, or could not be checked without following one, so Claude Code did not look inside it. Instead it write-denied ${WRITE_DENIED_PATHS.join(" and ")}, so git commands in the sandbox cannot update any linked worktree of this repository (add, commit, checkout), including this session's own.${GET_PLATFORM_FN()==="macos"?" On macOS this holds only if one of those paths is the real folder: a deny on a link does not stop writes through it.":""} If you did not make that link, remove it and restart Claude Code.
diff --git a/system-prompts/data-sandbox-project-false-value-ignored-under-higher-priority-policy.md b/system-prompts/data-sandbox-project-false-value-ignored-under-higher-priority-policy.md
new file mode 100644
index 0000000..9393a47
--- /dev/null
+++ b/system-prompts/data-sandbox-project-false-value-ignored-under-higher-priority-policy.md
@@ -0,0 +1,9 @@
+
+${HIGHER_PRIORITY_SANDBOX_POLICY_CONDITIONS}, and managed, --settings or user settings set true, false from ${PROJECT_SETTINGS_SOURCE_DESCRIPTION} is ignored (true there still applies).
diff --git a/system-prompts/data-sandbox-unchecked-protected-paths-warning.md b/system-prompts/data-sandbox-unchecked-protected-paths-warning.md
new file mode 100644
index 0000000..6a1459a
--- /dev/null
+++ b/system-prompts/data-sandbox-unchecked-protected-paths-warning.md
@@ -0,0 +1,8 @@
+
+[Sandbox] ${UNCHECKED_PROTECTED_PATHS.join(", ")} (among the paths Claude Code itself keeps from sandboxed commands) could not be checked without following a link that may lead to another computer, or could not be checked at all. While that is so, no sandbox filesystem grant from the project's settings files is applied: sandbox.filesystem.allowRead and allowWrite, additionalDirectories, and Edit allow rules as write roots, from .claude/settings.json and .claude/settings.local.json (a remembered /add-dir lands there), or from a Windows registry policy key the user can write. Grants from your user settings, administrator-managed settings and the command line still apply. If a link on that path is not one you or your administrator made, remove it; otherwise put the grants you need in your user settings.
diff --git a/system-prompts/data-sdk-api-error-kind-field.md b/system-prompts/data-sdk-api-error-kind-field.md
index 793bd59..04aa0d4 100644
--- a/system-prompts/data-sdk-api-error-kind-field.md
+++ b/system-prompts/data-sdk-api-error-kind-field.md
@@ -1,6 +1,6 @@
-@internal Typed kind of the API error when is_api_error_message is true, for consumers that key on the cause instead of the message text (the text stays the fallback and may change). Absent when the error has no typed kind; new values are added over time, so treat an unknown value as absent. dlp_request_denied: the request was refused by the organization's data-loss-prevention policy — terminal for that request body; a consumer replaying history must not re-send it. claude_code_version_too_old: the API refused the request because this Claude Code build is below a required minimum version — the selected model's minimum, or an organization-policy floor; only the message text says which and names the versions, and the remedy is updating Claude Code. safety_monitor_blocked: a server-side safety monitor blocked the model's response for this turn; terminal for the turn, so a consumer replaying history must not re-send the prompt that led to it. effort_requires_thinking: the model refuses the requested effort level while thinking is off; api_error_params.effort names the level. advisor_incompatible: the configured advisor model cannot be paired with this request's model. tool_history_mismatch: the API rejected the conversation's tool_use/tool_result pairing (a tool_use without its result, an orphaned tool_result, a duplicate tool_use id); re-sending the same history fails the same way, and rewinding the conversation is the remedy. autocompact_thrashing: the context refilled to the limit right after each automatic compaction, repeatedly, so the turn stopped; something in context is too large for the window. pdf_too_large: a PDF is over the API's page or size limit. pdf_password_protected: a PDF is encrypted. stored_login_refused: the stored claude.ai login was about to go out while the API key helper set by the machine's administrator or the organization's settings isn't in effect, or its settings couldn't be confirmed; no request was sent, and the message text says who can clear it. media_removed: the API refused an image or a document in the conversation, so Claude Code leaves it out of the requests that follow. The frame comes either before the answer of a retry that already left it out, or as the error that ended the turn (the result then repeats this value). api_error_params.media_reason says why, and api_error_params.media which kind of block. Other kinds (pdf_too_large, pdf_password_protected) and some untyped errors also leave the block out, so the absence of this value does not mean that nothing was left out. no_response: no response arrived before the first-byte deadline on any attempt, usually a proxy or gateway that holds responses until they complete. tls_untrusted_ca: the API's TLS certificate comes from an authority this process does not trust, usually a TLS-inspecting proxy or a gateway signed by a private CA. gateway_content_type: a proxy between Claude Code and Amazon Bedrock rewrote Bedrock's event-stream response. connection_failed: Claude Code could not reach the API host (DNS, a refused connection, no route, a proxy that refused the tunnel) or lost the connection (an open connection was cut, or the response body ended early). An error with a TLS error code never carries it: the untrusted-authority codes are tls_untrusted_ca, and the other certificate and handshake codes (for example an expired, revoked or wrong-host certificate) have no typed kind. A connection that is cut during the TLS handshake arrives with a socket code such as ECONNRESET and carries it. The client's own request timeout and a connection error whose cause Claude Code cannot tell have no typed kind either. stream_idle_timeout: Claude Code's own idle timer gave up waiting for more of the response on an open stream. It says that the timer fired, not that the server stalled: a long gap in a slow answer, a proxy that holds the response back, or a sleep that the timer cannot tell from a stall (a short one, for example) fires it too. computer_slept: a Claude Code timer found that the machine slept through its wait for the response. A claude-code-server session never sends it, because its process is not the person's computer; there the same failure has no typed kind. provider_credentials: a third-party cloud provider (api_error_params.provider) refused the request's credentials or access; api_error_params.remedy says what fixes it. gateway_signin_required: the Cloud gateway provider is configured and nobody is signed in to it; no request was sent. gateway_session_expired: the Cloud gateway session expired; no request was sent; api_error_params.remedy is gateway_token when the token came from the environment, which signing in does not replace. api_key_auth_disabled: the organization turned off API key authentication and the active credential is an API key. org_disabled_credential: the ANTHROPIC_API_KEY environment variable belongs to a disabled organization and overrides the account sign-in. invalid_credential_header: a credential or header value (from the environment, a helper, or the saved sign-in) cannot be sent as an HTTP header; caught before the request was sent. model_requires_usage_credits: the model runs on usage credits for this account, and credits are off or the included allowance is used up. long_context_credits_required: a request over the 200K-token context boundary needs usage credits. consent_unanswered: the model now uses usage credits and the prompt to confirm went unanswered; nothing was sent. no_allowed_fallback: switching off the model was needed (the usage-credits prompt was declined) and the model policy allows no other model. model_substitution_disabled: switching off the model was needed and CLAUDE_CODE_NO_MODEL_FALLBACK forbids it. field_not_granted: the organization is not allowed to send a field in the request, and the message text says what to do. usage_limit_reached: the request was refused by the account's own usage limit, a plan limit (session, weekly or per-model) or a spend cap on usage credits; api_error_params.rate_limit_info says which and when it resets. The other refusals that share error rate_limit never carry it: an organization policy that turned the service off (no limit was reached, and no reset clears it), a temporary server throttle, a capacity switch-off, an entitlement refusal shown in the server's own words, or a 429 on an API key or a third-party cloud provider. empty_completion: the model ended its turn with no visible text, its one retry had no content block, and a repeat of that retry had no visible text, on a person's own prompt with no reply and no tool result after it; the turn ends as an error so that a host shows a failure and not a blank reply. malformed_tool_use_exhausted: the model's reply stopped for a tool call and held none, and the loop's retries ran out; terminal for the turn, so a consumer replaying history must not re-send the prompt by itself. unattended_start_limit: a cloud session's process that starts where an earlier one died carries on what that one left unfinished; by the session's record, as many processes as are allowed (a number that is configured; with none there is no limit) had already done so within 24 hours with no message from a person in between, and this one did not; no request was sent; terminal for the turn, so a consumer replaying history must not re-send the prompt that led to it; anything new sent to the session runs as usual; a person's message allows it again, and so does time, as earlier starts pass out of the 24 hours. unattended_start_unconfirmed: a cloud session's process did not carry on what an earlier one left unfinished, because it could not check whether it was allowed to: the session's record of such starts could not be read or written, or the check failed; no request was sent; not terminal: a later process may carry the turn on. network_edge_blocked: the firewall in front of Anthropic's API refused the request for its network location (an unsupported region, or a VPN, proxy or network routing through one) before it reached the API; no sign-in or credential changes this.
+@internal Typed kind of the API error when is_api_error_message is true, for consumers that key on the cause instead of the message text (the text stays the fallback and may change). Absent when the error has no typed kind; new values are added over time, so treat an unknown value as absent. dlp_request_denied: the request was refused by the organization's data-loss-prevention policy — terminal for that request body; a consumer replaying history must not re-send it. claude_code_version_too_old: the API refused the request because this Claude Code build is below a required minimum version — the selected model's minimum, or an organization-policy floor; only the message text says which and names the versions, and the remedy is updating Claude Code. safety_monitor_blocked: a server-side safety monitor blocked the model's response for this turn; terminal for the turn, so a consumer replaying history must not re-send the prompt that led to it. effort_requires_thinking: the model refuses the requested effort level while thinking is off; api_error_params.effort names the level. advisor_incompatible: the configured advisor model cannot be paired with this request's model. tool_history_mismatch: the API rejected the conversation's tool_use/tool_result pairing (a tool_use without its result, an orphaned tool_result, a duplicate tool_use id); re-sending the same history fails the same way, and rewinding the conversation is the remedy. autocompact_thrashing: the context refilled to the limit right after each automatic compaction, repeatedly, so the turn stopped; something in context is too large for the window. pdf_too_large: a PDF is over the API's page or size limit. pdf_password_protected: a PDF is encrypted. stored_login_refused: the stored claude.ai login was about to go out while the API key helper set by the machine's administrator or the organization's settings isn't in effect, or its settings couldn't be confirmed; no request was sent, and the message text says who can clear it. media_removed: the API refused an image or a document in the conversation, so Claude Code leaves it out of the requests that follow. The frame comes either before the answer of a retry that already left it out, or as the error that ended the turn (the result then repeats this value). api_error_params.media_reason says why, and api_error_params.media which kind of block. Other kinds (pdf_too_large, pdf_password_protected) and some untyped errors also leave the block out, so the absence of this value does not mean that nothing was left out. no_response: no response arrived before the first-byte deadline on any attempt, usually a proxy or gateway that holds responses until they complete. tls_untrusted_ca: the API's TLS certificate comes from an authority this process does not trust, usually a TLS-inspecting proxy or a gateway signed by a private CA. gateway_content_type: a proxy between Claude Code and Amazon Bedrock rewrote Bedrock's event-stream response. connection_failed: Claude Code could not reach the API host (DNS, a refused connection, no route, a proxy that refused the tunnel) or lost the connection (an open connection was cut, or the response body ended early). An error with a TLS error code never carries it: the untrusted-authority codes are tls_untrusted_ca, and the other certificate and handshake codes (for example an expired, revoked or wrong-host certificate) have no typed kind. A connection that is cut during the TLS handshake arrives with a socket code such as ECONNRESET and carries it. The client's own request timeout and a connection error whose cause Claude Code cannot tell have no typed kind either. stream_idle_timeout: Claude Code's own idle timer gave up waiting for more of the response on an open stream. It says that the timer fired, not that the server stalled: a long gap in a slow answer, a proxy that holds the response back, or a sleep that the timer cannot tell from a stall (a short one, for example) fires it too. computer_slept: a Claude Code timer found that the machine slept through its wait for the response. A claude-code-server session never sends it, because its process is not the person's computer; there the same failure has no typed kind. provider_credentials: a third-party cloud provider (api_error_params.provider) refused the request's credentials or access; api_error_params.remedy says what fixes it. gateway_signin_required: the Cloud gateway provider is configured and nobody is signed in to it; no request was sent. gateway_session_expired: the Cloud gateway session expired; no request was sent; api_error_params.remedy is gateway_token when the token came from the environment, which signing in does not replace. api_key_auth_disabled: the organization turned off API key authentication and the active credential is an API key. org_disabled_credential: the ANTHROPIC_API_KEY environment variable belongs to a disabled organization and overrides the account sign-in. invalid_credential_header: a credential or header value (from the environment, a helper, or the saved sign-in) cannot be sent as an HTTP header; caught before the request was sent. model_requires_usage_credits: the model runs on usage credits for this account, and credits are off or the included allowance is used up. long_context_credits_required: a request over the 200K-token context boundary needs usage credits. consent_unanswered: the model now uses usage credits and the prompt to confirm went unanswered; nothing was sent. no_allowed_fallback: switching off the model was needed (the usage-credits prompt was declined) and the model policy allows no other model. model_substitution_disabled: switching off the model was needed and CLAUDE_CODE_NO_MODEL_FALLBACK forbids it. field_not_granted: the organization is not allowed to send a field in the request, and the message text says what to do. usage_limit_reached: the request was refused by the account's own usage limit, a plan limit (session, weekly or per-model) or a spend cap on usage credits; api_error_params.rate_limit_info says which and when it resets. The other refusals that share error rate_limit never carry it: an organization policy that turned the service off (no limit was reached, and no reset clears it), a temporary server throttle, a capacity switch-off, an entitlement refusal shown in the server's own words, or a 429 on an API key or a third-party cloud provider. empty_completion: the model ended its turn with no visible text, and so did its one retry and a repeat of that retry, on a person's own prompt with no reply and no tool result after it; the turn ends as an error so that a host shows a failure and not a blank reply. malformed_tool_use_exhausted: the model's reply stopped for a tool call and held none, and the loop's retries ran out; terminal for the turn, so a consumer replaying history must not re-send the prompt by itself. unattended_start_limit: a cloud session's process that starts where an earlier one died carries on what that one left unfinished; by the session's record, as many processes as are allowed (a number that is configured; with none there is no limit) had already done so within 24 hours with no message from a person in between, and this one did not; no request was sent; terminal for the turn, so a consumer replaying history must not re-send the prompt that led to it; anything new sent to the session runs as usual; a person's message allows it again, and so does time, as earlier starts pass out of the 24 hours. unattended_start_unconfirmed: a cloud session's process did not carry on what an earlier one left unfinished, because it could not check whether it was allowed to: the session's record of such starts could not be read or written, or the check failed; no request was sent; not terminal: a later process may carry the turn on. network_edge_blocked: the firewall in front of Anthropic's API refused the request for its network location (an unsupported region, or a VPN, proxy or network routing through one) before it reached the API; no sign-in or credential changes this.
diff --git a/system-prompts/data-sdk-api-error-type-field.md b/system-prompts/data-sdk-api-error-type-field.md
new file mode 100644
index 0000000..11ea746
--- /dev/null
+++ b/system-prompts/data-sdk-api-error-type-field.md
@@ -0,0 +1,6 @@
+
+@internal The error.type of the response body for this API error, such as rate_limit_error, overloaded_error or invalid_request_error, copied through when it is an identifier (^[a-z][a-z0-9_]{0,63}$) and dropped otherwise. Where it is present, a host can read the class here and need not work it out from error, api_error_status or the message text. The value is what the body says, and Claude Code does not check it. A gateway or a third-party provider can use other values (see api_error_origin). A first-party 503 with x-should-retry: false says overloaded_error though the server said not to retry, so the value alone does not say that a retry can help. Claude Code writes some bodies itself: a hook's plain refusal (hook_denied) has the type permission_error, and claude-code-server's own answers to a request that it must not or cannot send have the type api_error or invalid_request_error. When retries of a 529 run out, it is the type of the last response. Absent when the response had no body or the body had no type, for a connection error, on a frame that Claude Code builds itself, and on dlp_request_denied: if its body said overloaded_error, a host could send again a request that must not be sent again.
diff --git a/system-prompts/data-sdk-oauth-token-refresh-refused-call-field.md b/system-prompts/data-sdk-oauth-token-refresh-refused-call-field.md
new file mode 100644
index 0000000..9e8f17d
--- /dev/null
+++ b/system-prompts/data-sdk-oauth-token-refresh-refused-call-field.md
@@ -0,0 +1,6 @@
+
+The kind of call that the server refused. 'model': a call to the model, which fails at once if the host answers with the same token or none, so the host renews. 'connector': a request to a claude.ai connector. 'other': neither, at a call site that someone has checked: its failure on an unchanged token ends no turn, start or session. To 'connector' or 'other' the host may first ask the server whether the token that it holds is still good, and keep it: it then answers with that token, which is no fault to these two; if the CLI held another, it takes this one. The check and the answer must fit well inside the CLI's 30 s limit, because a model call that joined the ask shares a timeout and fails. Only with token_state 'refused'. Absent on a 'refused' ask: the call site sets none, or the CLI is older, and the host renews, as before the field: such an ask can come from a call whose failure ends a turn. A caller does not settle for the answer 'no new token' to an ask that wanted less than its own ('not_refused', then 'connector' or 'other', then no refused_call, then 'model'): it asks again, and the host judges that ask on its own. A host takes any other value as absent.
diff --git a/system-prompts/data-sdk-reload-plugins-refuse-if-busy-field.md b/system-prompts/data-sdk-reload-plugins-refuse-if-busy-field.md
new file mode 100644
index 0000000..889f076
--- /dev/null
+++ b/system-prompts/data-sdk-reload-plugins-refuse-if-busy-field.md
@@ -0,0 +1,6 @@
+
+When true, a request that finds the session busy reloads nothing and gets an error whose error_code is reload_plugins_busy; ask again once the session is idle. Busy: a turn is running or still ending, a message is queued for one, a permission prompt taken over from an earlier process is unanswered, background work that keeps the session at running (an agent, a workflow and the like) is live, or an in-process teammate is at work; a session that only waits for a background shell command is not busy. For a host whose request can arrive late, such as one that reaches the session over SSH: a reload changes the commands, agents, hooks and servers under a running turn, agent or teammate. Default false: the request is served in a turn too. Only `claude -p` reads it, and says so in system/init.capabilities (reload_plugins_refuse_if_busy_v1) to a host on its pipes; a CLI that does not say so may ignore the field.
diff --git a/system-prompts/data-sdk-startup-failure-reason-codes.md b/system-prompts/data-sdk-startup-failure-reason-codes.md
index 22495ca..7bf22a9 100644
--- a/system-prompts/data-sdk-startup-failure-reason-codes.md
+++ b/system-prompts/data-sdk-startup-failure-reason-codes.md
@@ -1,6 +1,6 @@
-Why Claude Code refused to start, so a host can offer the fix instead of a retry. org_pin_api_key_conflict: managed settings pin a first-party or Cloud gateway sign-in, and an Anthropic API key or auth token is configured instead. provider_not_allowed: managed settings list the API providers this machine may use (allowedProviders), and the session is set up for one that is not listed. org_verify_failed: the sign-in's organization could not be verified against the pin (network, or a revoked token). org_pin_mismatch: the sign-in belongs to an organization the pin does not allow. managed_settings_invalid: managed policy settings could not be read, the pin names no organization, or managed model settings (deniedModels, or an availableModels list matched exactly) block the default model and leave no allowed model to use instead. remote_settings_required_unavailable: managed settings the organization requires could not be loaded. gateway_signin_required: the Cloud gateway ended this sign-in. gateway_access_denied: the Cloud gateway refused managed settings for this account. proxy_invalid: a proxy setting is not a complete URL. temp_dir_unusable: the per-user temp directory is unsafe or could not be created. cwd_unavailable: the working directory was deleted, moved or cannot be read. shell_tool_missing: Windows has no shell tool: Git Bash is missing, and PowerShell is missing or turned off by CLAUDE_CODE_USE_POWERSHELL_TOOL. session_held_by_background: the conversation to resume or continue is running as a background session. worktree_resume_refused: the resume was refused because the session's worktree failed its safety checks or the resume was launched from inside it; errors says whether a re-run continues without the worktree. worktree_unverified: the session's worktree could not be verified right now; retrying may succeed. cli_version_too_old: this Claude Code version is below the minimum Anthropic requires. bypass_root: bypass permissions mode was requested while running as root. org_config_required_unavailable: the organization requires its policy limits and managed settings before a session starts, and they could not be loaded (network, proxy, or an Anthropic error); retrying may succeed. org_config_refused: the organization requires its policy limits and managed settings, and Anthropic refused them for this sign-in (expired or revoked session, or the organization does not allow Claude Code for this account); signing in again or an admin's change is the fix, not a retry. sandbox_required_unavailable: a working sandbox is required and it cannot run in this session, for example because bubblewrap is missing or the platform is not supported; make the sandbox available or have the requirement changed. initialize_not_first: Claude Code was started with --await-initialize, and its first stdin line was not the initialize control request (another message, or not valid JSON). Either the host wrote something else first, or a wrapper script between the host and Claude Code read or changed that line; the fix is in the host or the wrapper script, not a retry. config_file_invalid: a configuration file read at start-up is not valid JSON; errors names the file and, for the global configuration file, how to restore or repair it. The file is left unchanged, and every start without a terminal is refused until it is repaired, so a retry alone does not help. managed_settings_not_delivered: Claude Code was started with --managed-settings-in-initialize and could not use the managed settings in the initialize control request. Stdin ended or failed first, or its first line was too long; the request has no managedSettings object, that object could not be used, or the request's plugins field is malformed; or the launch does not read that request at start-up or also passes --managed-settings. errors says which. Claude Code starts no session without them; the fix is in the host or in a wrapper script between it and Claude Code, not a retry.
+Why Claude Code refused to start, so a host can offer the fix instead of a retry. org_pin_api_key_conflict: managed settings pin a first-party or Cloud gateway sign-in, and an Anthropic API key or auth token is configured instead. provider_not_allowed: managed settings list the API providers this machine may use (allowedProviders), and the session is set up for one that is not listed. org_verify_failed: the sign-in's organization could not be verified against the pin (network, or a revoked token). org_pin_mismatch: the sign-in belongs to an organization the pin does not allow. managed_settings_invalid: managed policy settings could not be read, the pin names no organization, or managed model settings (deniedModels, or an availableModels list matched exactly) block the default model and leave no allowed model to use instead. remote_settings_required_unavailable: managed settings the organization requires could not be loaded. gateway_signin_required: the Cloud gateway ended this sign-in. gateway_access_denied: the Cloud gateway refused managed settings for this account. proxy_invalid: a proxy setting is not a complete URL. temp_dir_unusable: the per-user temp directory is unsafe or could not be created. cwd_unavailable: the working directory was deleted, moved or cannot be read. shell_tool_missing: Windows has no shell tool: Git Bash is missing, and PowerShell is missing or turned off by CLAUDE_CODE_USE_POWERSHELL_TOOL. session_held_by_background: the conversation to resume or continue is running as a background session. worktree_resume_refused: the resume was refused because the session's worktree failed its safety checks or the resume was launched from inside it; errors says whether a re-run continues without the worktree. worktree_unverified: the session's worktree could not be verified right now; retrying may succeed. cli_version_too_old: this Claude Code version is below the minimum Anthropic requires. bypass_root: bypass permissions mode was requested while running as root. org_config_required_unavailable: the organization requires its policy limits and managed settings before a session starts, and they could not be loaded (network, proxy, or an Anthropic error); retrying may succeed. org_config_refused: the organization requires its policy limits and managed settings, and Anthropic refused them for this sign-in (expired or revoked session, or the organization does not allow Claude Code for this account); signing in again or an admin's change is the fix, not a retry. sandbox_required_unavailable: a working sandbox is required and it cannot run in this session, for example because bubblewrap is missing or the platform is not supported; make the sandbox available or have the requirement changed. initialize_not_first: Claude Code was started with --await-initialize, and its first stdin line was not the initialize control request (another message, or not valid JSON). Either the host wrote something else first, or a wrapper script between the host and Claude Code read or changed that line; the fix is in the host or the wrapper script, not a retry. config_file_invalid: a configuration file read at start-up is not valid JSON; errors names the file and, for the global configuration file, how to restore or repair it, and invalid_config_file has the paths as fields. The file is left unchanged, and every start without a terminal is refused until it is repaired, so a retry alone does not help. managed_settings_not_delivered: Claude Code was started with --managed-settings-in-initialize and could not use the managed settings in the initialize control request. Stdin ended or failed first, or its first line was too long; the request has no managedSettings object, that object could not be used, or the request's plugins field is malformed; or the launch does not read that request at start-up or also passes --managed-settings. errors says which. Claude Code starts no session without them; the fix is in the host or in a wrapper script between it and Claude Code, not a retry.
diff --git a/system-prompts/data-sdk-stream-event-owed-rows-field.md b/system-prompts/data-sdk-stream-event-owed-rows-field.md
index ac80ebd..331177f 100644
--- a/system-prompts/data-sdk-stream-event-owed-rows-field.md
+++ b/system-prompts/data-sdk-stream-event-owed-rows-field.md
@@ -1,6 +1,6 @@
-@internal The uuids of the `project_doc_reads` rows that this turn wrote ahead of its first model request and that give the text of an added document. Such a row records Claude Code's own read of a claude.ai Project's documents. A document counts as added when no earlier row of the conversation gives a text under its name, or a later one of them tells the document of that name as removed. A row with only replaced, edited or removed documents is not listed, and neither is a row from before the conversation's first reply. Empty when the turn wrote no row to list. A consumer that read the transcript at the turn's start can tell from it whether that read lacks such a row. The rows may not have reached the session server yet. Present exactly when user_message_uuid is, on the same stream events; absent from older producers, and then it states nothing.
+@internal The uuids of the rows that this turn wrote ahead of its first model request and that change what Claude holds of the chat's claude.ai Project. A read of the transcript at the turn's start can lack them. A `project_doc_reads` row records Claude Code's own read of the Project's documents. It is listed when it names a document, and the conversation already held a reply or an earlier such row. A `session_context` row is listed when it replaces an earlier one and its `attachedProject` differs from that row's. Empty when the turn wrote no row to list. The rows may not have reached the session server yet. Present exactly when user_message_uuid is, on the same stream events; absent from older producers, and then it states nothing.
diff --git a/system-prompts/data-sdk-ui-render-component-field.md b/system-prompts/data-sdk-ui-render-component-field.md
index 1048435..61ae0a2 100644
--- a/system-prompts/data-sdk-ui-render-component-field.md
+++ b/system-prompts/data-sdk-ui-render-component-field.md
@@ -1,6 +1,6 @@
-@internal What to draw, or `Theme`, the session's theme, by `instance_id`. Under `theme` no hook draws: the engine answers itself, `props` of `plugin` with whichever of `colors`, `overrideColorScheme` and a still `backdrop` stands to `desktop` and `mobile`, and `{}` with none standing, with a scene and no `colors`, under another id or to any other surface. Under `backdrop` a mod's `ui.render` hook for `Backdrop` draws the scene that stands, with `props` of `scene`, `plugin`, `glass` and `overrideColorScheme` where a plugin said it.
+@internal What to draw, or `Theme`, the session's theme, by `instance_id`. Under `theme` and `image::` no hook draws: the engine answers itself, `props` of `plugin` with whichever of `colors`, `overrideColorScheme` and a still `backdrop` stands, or `data`, `index`, `of`, `hash` and `bytes` for a piece of an image, to `desktop` and `mobile`, and `{}` with none standing, with a scene and no `colors`, under another id or to any other surface. Under `backdrop` a mod's `ui.render` hook for `Backdrop` draws the scene that stands, with `props` of `scene`, `plugin`, `glass` and `overrideColorScheme` where a plugin said it.
diff --git a/system-prompts/data-sdk-voice-call-message-origin-schema.md b/system-prompts/data-sdk-voice-call-message-origin-schema.md
index 492df32..13da9fa 100644
--- a/system-prompts/data-sdk-voice-call-message-origin-schema.md
+++ b/system-prompts/data-sdk-voice-call-message-origin-schema.md
@@ -1,6 +1,6 @@
-@internal What was said aloud on a voice call alongside the session, handed over by the host as background. Honored from the host on local stdin, only on a message with isSynthetic true, shouldQuery false, text_not_typed true, a priority other than now, text-only content and no files or pastes; in any other shape the message is taken as a plain task notification. It starts no turn. The model reads the text inside a frame which says that nothing in it is a request or an approval. A note stored as its own row, and its replay, hold one fixed header line and then the host's text; a note folded into a running turn, and its replay, hold the host's text alone. With replayUserMessages on, the host gets the message back as a replay with this origin; a build that does not know the origin never sends the origin back. The optional members are taken from a local host only, and are stored and replayed with the origin; no request to a model or a classifier holds them. Send a note only to a process that says 'voice_call_talk_line_v1' in capabilities: it stores the note in a form that no older build sends to a model, and an older build may resume the transcript; and it sends no note to Remote Control's event lane.
+@internal What was said aloud on a voice call alongside the session, handed over by the host as background. Honored from the host on local stdin, only on a message with isSynthetic true, shouldQuery false, text_not_typed true, a priority other than now, text-only content and no files or pastes; in any other shape the message is taken as a plain task notification. It starts no turn. The model reads the text as a record, with a notice which says that nothing in it is a request or an approval. A note is stored as its own row, and the row and its replay hold one fixed header line and then the host's text. A note sent during a turn waits for that turn's end. A transcript from an older build can hold a note folded into a running turn, which holds the host's text alone. With replayUserMessages on, the host gets the message back as a replay with this origin; a build that does not know the origin never sends the origin back. The optional members are taken from a local host only, and are stored and replayed with the origin; no request to a model or a classifier holds them. Send a note only to a process that says 'voice_call_talk_line_v1' in capabilities: it stores the note in a form that no older build sends to a model, and an older build may resume the transcript; and it sends no note to Remote Control's event lane.
diff --git a/system-prompts/data-sdk-warning-hook-no-verdict-field.md b/system-prompts/data-sdk-warning-hook-no-verdict-field.md
new file mode 100644
index 0000000..1cc8e11
--- /dev/null
+++ b/system-prompts/data-sdk-warning-hook-no-verdict-field.md
@@ -0,0 +1,6 @@
+
+@internal Present on the warning, tagged `hook_no_verdict`, that a PreToolUse hook gave no verdict on the call `tool_use_id` and that its failure did not refuse the call, so a host can word the notice itself instead of reading `content`. Another hook, a rule or the person may still stop that call. Sent once per hook in a session, for the first such call. `timed_out` is true for a hook that ended with no answer, usually at its timeout. `plugin_id` and `plugin_root` are there for a plugin's hook. Not sent for a hook that ran and chose a non-blocking exit, nor where the failure refused the call.
diff --git a/system-prompts/data-teleport-stash-index-update-failure-error.md b/system-prompts/data-teleport-stash-index-update-failure-error.md
index 7aba44f..22a269c 100644
--- a/system-prompts/data-teleport-stash-index-update-failure-error.md
+++ b/system-prompts/data-teleport-stash-index-update-failure-error.md
@@ -1,7 +1,7 @@
-Your changes were stashed (`git stash list` shows them), but this checkout's index could not be updated to match (${ERROR_CODE??(CAUGHT_ERROR instanceof INDEX_INSTALL_ERROR_CLASS?CAUGHT_ERROR.message:`the ${INDEX_INSTALL_STEP_LABELS[INDEX_INSTALL_STEP]} failed`)}), so `git status` will show changes that are not there. First check that .git/index is a plain file: if it is a link, delete it, because git would write to the file it leads to. Then run `git reset` in this checkout (if that fails, check that .git/index is not a folder and that you can write to the .git folder; if git says index.lock exists and no other git is running, delete that file first) and run /teleport again; your changes stay in the stash.
+this checkout's index could not be updated to match (${ERROR_CODE??(CAUGHT_ERROR instanceof INDEX_INSTALL_ERROR_CLASS?CAUGHT_ERROR.message:`the ${INDEX_INSTALL_STEP_LABELS[INDEX_INSTALL_STEP]} failed`)}), so `git status` will show changes that are not there. First check that .git/index is a plain file: if it is a link, delete it, because git would write to the file it leads to. Then run `git reset` in this checkout (if that fails, check that .git/index is not a folder and that you can write to the .git folder; if git says index.lock exists and no other git is running, delete that file first)
diff --git a/system-prompts/data-teleport-stash-merge-autostash-refusal.md b/system-prompts/data-teleport-stash-merge-autostash-refusal.md
new file mode 100644
index 0000000..1448926
--- /dev/null
+++ b/system-prompts/data-teleport-stash-merge-autostash-refusal.md
@@ -0,0 +1,8 @@
+
+Nothing was stashed: this checkout has a MERGE_AUTOSTASH record. A stash would add the commit it names to your stash list, where `git stash pop` would apply it. If you started a merge that set your changes aside, finish or abort that merge. If you did not, a sandboxed command may have planted the record: delete ${MERGE_AUTOSTASH_PATH} if it exists, then run `git update-ref -d --no-deref MERGE_AUTOSTASH`. Neither applies anything. A branch or tag of that name causes this too: rename it. Then run /teleport again.
diff --git a/system-prompts/data-teleport-stash-ref-without-entries-refusal.md b/system-prompts/data-teleport-stash-ref-without-entries-refusal.md
new file mode 100644
index 0000000..3f31c8c
--- /dev/null
+++ b/system-prompts/data-teleport-stash-ref-without-entries-refusal.md
@@ -0,0 +1,9 @@
+
+Nothing was stashed: refs/stash is set, but git lists no stash entries, so /teleport could not check what a stash adds. A sandboxed command may have caused this, so do not run `git stash pop` yet. Run `${STASH_SHOW_COMMAND} | cat -v` (${GIT_BASH_DISPLAY_NOTE}). If it shows changes you did not make, run `git stash drop` and look again. If it says refs/stash@{0} is not a valid reference, run `git update-ref -d refs/stash`; `git fsck --lost-found` then lists that commit as dangling. If it prints any other error, or shows only your own changes, run `git stash --include-untracked`, which clears this. Then run /teleport again.
diff --git a/system-prompts/data-teleport-stash-unexpected-entry-refusal.md b/system-prompts/data-teleport-stash-unexpected-entry-refusal.md
new file mode 100644
index 0000000..b91b9cf
--- /dev/null
+++ b/system-prompts/data-teleport-stash-unexpected-entry-refusal.md
@@ -0,0 +1,12 @@
+
+the stash list holds something /teleport did not put there, or could not be read, so /teleport stopped. ${UNSAFE_GIT_DIR_ENTRY===null?"":`Before any git stash command: ${TRUNCATE_PATH_FN(REPLACE_CONTROL_CHARACTERS_FN(UNSAFE_GIT_DIR_ENTRY),200)} (or a folder on the way to it) is a link, names another ref, cannot be read, or is not a plain file, and git would read and write through it. If you did not set this up, remove the link, or a stash file that reads "ref: …" (the file a link leads to is not touched). Never delete HEAD itself, and at the object store or the ref table first move what a link leads to back into a plain folder at that name, since it may be your repository's own objects or refs. `}Do not run `git stash pop` yet: a sandboxed command may have planted an entry through a MERGE_AUTOSTASH record, and `git stash list` can hide that entry or show it under this stash's name. Run `${STASH_SHOW_COMMAND} | cat -v` (${GIT_BASH_DISPLAY_NOTE}), which shows the changes pop would make. If they are the changes you just made, pop. If they are an older stash of yours, or git finds no stash, leave it: this stash's entry is gone. If they are anything else, run `git stash drop`, which drops that entry, and run that `git stash show` again.
diff --git a/system-prompts/data-tool-result-display-metadata-field.md b/system-prompts/data-tool-result-display-metadata-field.md
index 968d3a4..dfac5d3 100644
--- a/system-prompts/data-tool-result-display-metadata-field.md
+++ b/system-prompts/data-tool-result-display-metadata-field.md
@@ -1,6 +1,6 @@
-@internal Display metadata for this message's tool_result blocks, keyed by tool_use_id. non_execution_kind is the harness-stamped reason an is_error:true result did not carry the tool's own execution output (user-rejected / permission-rule / automode-* / interrupted / cancelled); absent means the tool ran to completion. remedy is the fix a host can offer for what the result reports (see SDKToolResultRemedy). An entry may carry a remedy alone, on a result that did run (a staged write, a sandbox violation on exit 0), so whether a call ran is read from non_execution_kind, never from an entry's presence. built_in_tool is the name of the Claude Code built-in tool whose own completed run, in the process that wrote this message, made the result, stamped for the few built-ins a host acts on; it is absent once a hook rewrote the result, on a message that the process read from a session file or took from another frame (a resumed session), on the result of a call that a client wrote (a handed-over turn), and on a subagent's messages. History that the same process sends again (historical: true) still carries it. On a subagent's messages (parent_tool_use_id set), entries carry only permission_decision, and beside a reject also non_execution_kind, unless a run had reached the tool before the refusal. So on those messages an absent non_execution_kind does not say that the call ran: a call that was cut has none, and older builds send none. permission_decision is how the call's permission check ended, as Claude Code's claude_code.tool.blocked_on_user trace span reports it. decision is accept, reject or cancelled. source names what decided: config, classifier, hook, user_temporary, user_permanent, user_reject, user_abort, or invalid_input, permission_updated_input_invalid or server_fallback_tombstone when bad input or the server ended the wait. Which label a rule gets depends on the session: in SDK sessions a rule for this session or in the user's own settings reports user_temporary, user_permanent or user_reject and other rules report config, while interactive sessions report every rule as config. reason_type is the kind of reason the check gave when it decided by itself, such as 'rule' whenever a rule decided, or 'classifier' for an auto-mode allow, which reports source config. reason_type is absent when a person or the SDK host answered the question, when the check's outcome was still a question, and for user_abort and the last three sources above. auto_mode_rule rides beside a reason_type of 'classifier' when the server-side auto-mode classifier named one of the session's own settings.autoMode entries as deciding the call: list is allow (an allow exception, on an accept) or soft_deny (on a reject), and name is that entry's first line, without escape sequences or hidden characters, up to a colon followed by whitespace or the line's end, or up to a space and '[', trimmed and at most 120 characters. The name's text is the session's own, never classifier output, but which entry is named is the classifier's claim. permission_decision is absent when the call ended before its permission check, and on some results built apart from the call's own run, such as a plugin hook's refusal, a hook's rewrite of a denied result, or the result the tool executor writes for a queued call that a stop cancelled. user_feedback is the deny comment a human typed at a permission prompt (the local dialog or a bridge client's deny UI), before the CLI's reject-message framing — absent when the deny was resolved by a coordinator, channel relay, or hook rather than a human. A deny answered through the SDK control protocol (can_use_tool / --permission-prompt-tool) currently reports as 'permission-rule' with no user_feedback, because that wire carries no human-provenance signal. Wrapper-level sibling — never inside `message.content` — so it is not replayed to the model. Clients read this instead of string-matching the result prose.
+@internal Display metadata for this message's tool_result blocks, keyed by tool_use_id. non_execution_kind is the harness-stamped reason an is_error:true result did not carry the tool's own execution output (user-rejected / permission-rule / automode-* / interrupted / cancelled); absent means the tool ran to completion. remedy is the fix a host can offer for what the result reports (see SDKToolResultRemedy). An entry may carry a remedy alone, on a result that did run (a staged write, a sandbox violation on exit 0), so whether a call ran is read from non_execution_kind, never from an entry's presence. built_in_tool is the name of the Claude Code built-in tool whose own completed run, in the process that wrote this message, made the result, stamped for the few built-ins a host acts on; it is absent once a hook rewrote the result, where a hook or a permission answer changed the call's input, on a message that the process read from a session file or took from another frame (a resumed session), on the result of a call that a client wrote (a handed-over turn), and on a subagent's messages. History that the same process sends again (historical: true) still carries it. run_input stands only beside built_in_tool, and only from a worker that takes its messages over Claude Code Remote. Its one value today, person_typed, says that a person typed, in the iOS, Android, claude.ai or desktop app, every message that opened the run under way or joined it up to this call, as far as the message itself shows: it came from such an app, with no mark that an app, the server or another session composed it. What a typed command or skill expands to, what hooks add and what tools return is not judged. On any doubt it is absent, and today more often than that meaning requires: also where the person's own message joined the run. A host reads absence, and any value that it does not know, as no. On a subagent's messages (parent_tool_use_id set), entries carry only permission_decision, and beside a reject also non_execution_kind, unless a run had reached the tool before the refusal. So on those messages an absent non_execution_kind does not say that the call ran: a call that was cut has none, and older builds send none. permission_decision is how the call's permission check ended, as Claude Code's claude_code.tool.blocked_on_user trace span reports it. decision is accept, reject or cancelled. source names what decided: config, classifier, hook, user_temporary, user_permanent, user_reject, user_abort, or invalid_input, permission_updated_input_invalid or server_fallback_tombstone when bad input or the server ended the wait. Which label a rule gets depends on the session: in SDK sessions a rule for this session or in the user's own settings reports user_temporary, user_permanent or user_reject and other rules report config, while interactive sessions report every rule as config. reason_type is the kind of reason the check gave when it decided by itself, such as 'rule' whenever a rule decided, or 'classifier' for an auto-mode allow, which reports source config. reason_type is absent when a person or the SDK host answered the question, when the check's outcome was still a question, and for user_abort and the last three sources above. auto_mode_rule rides beside a reason_type of 'classifier' when the server-side auto-mode classifier named one of the session's own settings.autoMode entries as deciding the call: list is allow (an allow exception, on an accept) or soft_deny (on a reject), and name is that entry's first line, without escape sequences or hidden characters, up to a colon followed by whitespace or the line's end, or up to a space and '[', trimmed and at most 120 characters. The name's text is the session's own, never classifier output, but which entry is named is the classifier's claim. permission_decision is absent when the call ended before its permission check, and on some results built apart from the call's own run, such as a plugin hook's refusal, a hook's rewrite of a denied result, or the result the tool executor writes for a queued call that a stop cancelled. user_feedback is the deny comment a human typed at a permission prompt (the local dialog or a bridge client's deny UI), before the CLI's reject-message framing — absent when the deny was resolved by a coordinator, channel relay, or hook rather than a human. A deny answered through the SDK control protocol (can_use_tool / --permission-prompt-tool) currently reports as 'permission-rule' with no user_feedback, because that wire carries no human-provenance signal. Wrapper-level sibling — never inside `message.content` — so it is not replayed to the model. Clients read this instead of string-matching the result prose.
diff --git a/system-prompts/data-wrapup-tool-result-nextstep-field.md b/system-prompts/data-wrapup-tool-result-nextstep-field.md
index 31bb2bb..6612f39 100644
--- a/system-prompts/data-wrapup-tool-result-nextstep-field.md
+++ b/system-prompts/data-wrapup-tool-result-nextstep-field.md
@@ -1,8 +1,8 @@
-The next-step box for this call, to draw without reading the tool's input. Open the box when blocked is true or actions isn't empty; otherwise draw no box. Trust it only when the result's entry in tool_result_meta has built_in_tool set to this tool. It is built from the input the call ran with, which a permission answer or a hook before the call may have rewritten, so built_in_tool does not say that Claude wrote it. The recap and the question are each cleaned to one line of at most ${MAX_NEXT_STEP_TEXT_LENGTH} UTF-16 units, ending in an ellipsis where that limit cut it, with no whole markers, no terminal escape sequences and no invisible characters other than those an emoji or a word needs. A placeholder such as [REDACTED] or stands in place of most keys and tokens recognised by their shape. A private key that could not be masked, or a tool call's own markup, is cut with all the text after it, and an ellipsis then shows only if what is left still passes that limit or the text ran past ${4*MAX_NEXT_STEP_TEXT_LENGTH} units. Draw the recap and the question as plain text, as they stand: they may hold Markdown, HTML or links, and normalizing them can undo the cleaning. Absent from a result stored by an earlier version of Claude Code.
+The next-step box for this call, to draw without reading the tool's input. Open the box when blocked is true or actions isn't empty; otherwise draw no box. Trust it only when the result's entry in tool_result_meta has built_in_tool set to this tool. It is built from the input the call ran with; where a permission answer or a hook before the call changed that input, the entry has no built_in_tool. The recap and the question are each cleaned to one line of at most ${MAX_NEXT_STEP_TEXT_LENGTH} UTF-16 units, ending in an ellipsis where that limit cut it, with no whole markers, no terminal escape sequences and no invisible characters other than those an emoji or a word needs. A placeholder such as [REDACTED] or stands in place of most keys and tokens recognised by their shape. A private key that could not be masked, or a tool call's own markup, is cut with all the text after it, and an ellipsis then shows only if what is left still passes that limit or the text ran past ${4*MAX_NEXT_STEP_TEXT_LENGTH} units. Draw the recap and the question as plain text, as they stand: they may hold Markdown, HTML or links, and normalizing them can undo the cleaning. Absent from a result stored by an earlier version of Claude Code.
diff --git a/system-prompts/skill-artifact-design-direction-cards.md b/system-prompts/skill-artifact-design-direction-cards.md
new file mode 100644
index 0000000..6147520
--- /dev/null
+++ b/system-prompts/skill-artifact-design-direction-cards.md
@@ -0,0 +1,30 @@
+
+- Swiss timetable. Type: Archivo width 112 wt 800 + Archivo. Colour: white #fff, black, signal red #e4002b, one grey; night #0e0e0e. Layout: strict left edge, huge numerals hanging in a wide left column, thick top rule.
+- Survey map. Type: Overpass 800 + Overpass. Colour: sea #bfe3ee, land #fbfbf2, contour brown #a8551b, ink #13242e, route magenta #d6127e; night #0d1c24. Layout: picture full bleed, text in a margin panel like a map legend.
+- Annual report cover. Type: Sofia Sans Condensed 800 + Instrument Sans. Colour: white, black, vermilion #ff4a1c as one solid chart mass; night vermilion on #120d0b. Layout: the chart is a giant flat shape bleeding off two edges, headline sits in its empty corner.
+- Scientific plate. Type: Literata 600 display size + Literata. Colour: white, ink #10131a, two inks: carmine #c1121f and prussian #0b4f8a; night #0f1217. Layout: large figure with leader-line callouts, narrow text column at right.
+- Newspaper graphics desk. Type: Libre Franklin 800 + Libre Franklin. Colour: pale pink #ffe3dc ground, ink #1a1817, teal #0d7680, claret #990f3d; night #1c1512. Layout: headline finding top left, one wide annotated chart, small multiples beneath.
+- Pharmacy label. Type: Schibsted Grotesk 800 + Schibsted Grotesk. Colour: white, one flat block each of emerald #00a86b and pale pink #ffd6e0, black; night #04140d. Layout: page cut into two or three rectangles like a medicine box.
+- Metro diagram. Type: Hammersmith One + Cabin. Colour: white, line colours red #e6261f blue #0a5cff amber #ffb400 green #00a05a, black; night #0e1116. Layout: one thick line with 45-degree bends organises the whole page.
+- Control room. Type: Saira Semi Condensed 600 + Saira. Colour: steel #d5dbe0 panel, black, white, alarm red #e8251a; night #16191c. Layout: one dominant readout at 40% width, long thin strips stacked beside it.
+- Architect monograph. Type: Hanken Grotesk 500 + Hanken Grotesk. Colour: white, black, concrete #d8d8d4 planes, safety yellow #ffe600 on one element; night #1a1a19. Layout: title modest in a corner, wide margins, one big exact drawing.
+- Weather chart. Type: Red Hat Display 800 + Red Hat Text. Colour: navy #0b1433 ground in both themes; scale from #3b4fd8 through #4fc1c6, #f7e463 to #e8452c. Layout: full-bleed field of colour-coded cells or isobars, legend as a long bar.
+- Product changelog. Type: Space Grotesk 500 + Inter. Colour: white, ink #0b0b10, signal green #00a862 as a full band behind the main figure, mist #eafaf2; night #07130d. Layout: headline across the full width, chart beneath at full bleed, figures in one ruled row.
+- Night desk. Type: Inter Tight 300 very large + Inter. Colour: plum-black #1a0f1f ground in both themes, white, rose #ff8fa3, dusk #8f7aa8. Layout: chart fills the screen, figures float in its empty corner.
+- Monochrome plus one. Type: Space Grotesk 700 + Inter. Colour: black and white only, plus one field of mint #b8f5d0; night black, mint as ink. Layout: half the screen is the colour field holding the number, half is white holding the chart.
+- Weekly leader page. Type: Source Serif 4 600 display + Inter. Colour: white, ink #121212, red #e3120b, slate #2e6f8e, panels of pale #e3ecf0; night #10191e. Layout: red tab at top left, headline full width, panels divided by rules on one baseline.
+- Atlas plate. Type: Hanken Grotesk 300 very large + Hanken Grotesk. Colour: deep sea #0a2540 ground in both themes, sky #8fd3ff, white, signal pink #ff5c8a. Layout: three horizontal strips on one shared axis, top one tallest.
+- Precision instrument. Type: Geist 500 + Geist. Colour: aluminium #e9eaec, black, orange #ff4f00 pointer; night #141517. Layout: one long horizontal scale across the screen, readout above it, controls below.
+- Sports federation report. Type: Outfit 600 + Inter. Colour: one field of sky #4fb3e8, white, ink #10223a, orange #ff7a1a for the point that matters; night #0b1726. Layout: colour band on top with finding and figures, small multiples in a row beneath.
+- Central bank bulletin. Type: Libre Caslon Text 400 large + Inter. Colour: white, ink #14161a, one deep teal #0b5d63, rule grey; night #0d1416. Layout: title across the top, two-thirds chart, one-third ruled column of figures.
+- Engineering datasheet. Type: Space Grotesk 600 + Inter. Colour: cobalt #1240e0 title strip, white, black, pale #eef2ff zones; night #0a0d18. Layout: narrow title strip, dense grid of exact panels sharing one baseline.
+- Journal article. Type: Spectral 500 + Inter. Colour: white, ink #14110f, ultramarine #1d2bd8, cool grey; night #0d0e18. Layout: one reading column with wide figures breaking out to the right, notes in the margin.
+- Transit authority. Type: Inter Tight 700 + Inter. Colour: deep green #00463c ground for the first screen, white, signal lime-white #e6f5c8, coral #ff6f59 for trouble; white below. Layout: finding across the top, wide chart, ruled figure row at the foot.
+- Audit report. Type: Manrope 700 + Manrope. Colour: white, ink #101418, one slab of warm yellow #ffd84a behind the key figure only, slate; night #101418. Layout: figure-first: chart at the very top, finding set beneath it, table follows.
+- Lab poster. Type: Red Hat Display 700 + Red Hat Text. Colour: oxford blue #0b1f4b band, white panels, magenta #e0115f for the result; night #070f26. Layout: small multiples in a row sharing one scale, title band above.
+- Planning office. Type: Schibsted Grotesk 700 + Schibsted Grotesk. Colour: stone #e7e4df, black, brick #c8371e, white panels; night #171513. Layout: table-led: a large, beautifully ruled table is the picture, verdict in the margin.
+- Ice and ink. Type: Space Grotesk 500 + Inter. Colour: ice #dff1f7 ground, ink #06222e, one hot pink #ff2e7e; night #06222e. Layout: chart left bleeding off the edge, finding right, figures below.
diff --git a/system-prompts/skill-artifact-design-page-contract.md b/system-prompts/skill-artifact-design-page-contract.md
new file mode 100644
index 0000000..c342a7f
--- /dev/null
+++ b/system-prompts/skill-artifact-design-page-contract.md
@@ -0,0 +1,20 @@
+
+# Page contract
+
+Author HTML (a .md file only when a loaded skill says so). The file is wrapped in a doctype/html/head/body skeleton when published (charset, viewport-fit=cover, color-scheme light, zero body margin, 14px system font, safe-area padding on :root, img{max-width:100%}, [hidden]{display:none!important}, so toggle with el.hidden, not style.display), so write only the content: , , font links,