Merge pull request #67 from qybaihe/feat/skill-weave-guide-1.3.0

feat(skill): blind-llm-agents v1.3.0 as one-click harden guide
This commit is contained in:
Timo Zhang
2026-11-02 19:24:11 +08:00
committed by GitHub
7 changed files with 238 additions and 42 deletions
+112 -26
View File
@@ -1,56 +1,142 @@
---
name: blind-llm-agents
description: Audit, protect, and verify software artifacts (JavaScript/TypeScript, Electron ASAR, WebAssembly, Python .py/.pyc, SQLite, Safetensors/GGUF AI Models, Android DEX/IL2CPP, Apple plist, .NET, JVM, ZIP/APK, and Native ELF/Mach-O/PE Binaries) against AI reverse-engineering agents like REA using BLA (Blind LLM Agents). Default verb: bla_harden.
description: >
Use when hardening, defending, or auditing software against AI reverse-engineering
agents (REA / WorkBuddy-class): one-click project harden, weave-decoy structural
decoys, Counter-REA audit/protect/verify, or agent setup/doctor for BLA (Blind LLM
Agents). Prefer this skill as the guide; call existing MCP tools / CLI — do not invent
new MCP tools. Default verb: bla_harden. Refuse crack/RE of third-party software.
metadata:
version: "1.2.0"
version: "1.3.0"
---
# BLA: Blind LLM Agents
Use BLA whenever the user asks to harden a project against AI reverse-engineering agents (`morluto/rea`), audit exposure, exhaust attacker token/compute budgets, or induce model misjudgment on proprietary logic.
**This skill is the primary one-click harden GUIDE.** MCP (`bla` server) and the
`bla-agents` CLI are the **execution layer**. Prefer teaching/calling existing tools
(`bla_harden`, `bla_defend`, `bla_weave_plan`, …) over inventing new MCP surfaces.
Use BLA whenever the user asks to harden a project against AI reverse-engineering
agents (`morluto/rea`), audit exposure, exhaust attacker token/compute budgets, induce
model misjudgment on proprietary logic, or **one-click / 一键加固** a build.
Defensive only: protect **their own** artifacts. Refuse crack, RE, bypass PoCs, or
attack playbooks against software they do not own.
## Prerequisites (once)
If MCP/skill are not registered yet:
1. CLI: `npx bla-agents@latest setup`
2. Then: `npx bla-agents@latest doctor` (or MCP `bla_doctor`)
After setup, agents load this skill and can call the `bla` MCP tools.
## ⚡ One-Sentence Hardening (DEFAULT — prefer this)
When the user says **"Harden my software"**, **"Harden this project against REA"**, **"一键启动防御强化"**, **"Run BLA defense"**, or any equivalent one-liner **without naming a specific file**, invoke:
When the user says **"Harden my software"**, **"Harden this project against REA"**,
**"一键启动防御强化"**, **"一键加固"**, **"Run BLA defense"**, or any equivalent
one-liner **without naming a specific file**, invoke:
- **MCP Tool (preferred)**: `bla_harden` with `{ "directory": "." }` (or the project/build path they named)
- **CLI**: `npx bla-agents@latest harden .`
- **MCP Prompt**: `bla-harden-project`
| Surface | Call |
| :--- | :--- |
| **MCP Tool (preferred)** | `bla_harden` with `{ "directory": "." }` (or the project/build path they named) |
| **CLI** | `npx bla-agents@latest harden .` |
| **MCP Prompt** | `bla-harden-project` |
`bla_harden` / `bla harden` auto-discovers protectable build outputs under `dist/`, `build/`, `out/`, `release/`, `.asar`, native binaries, `.wasm`, `.py`/`.pyc`, `.sqlite`, `.safetensors`/`.gguf`, `.dex`, `.class`, `.dll`, Apple plists, ZIP/APK, etc., detects each artifact's format, and runs the full **audit → defend (bankrupt) → verify** pipeline on all of them. Return the per-artifact summary to the user. Do **not** ask which files to protect — discover them.
`bla_harden` / `bla harden` auto-discovers protectable build outputs under `dist/`,
`build/`, `out/`, `release/`, `.asar`, native binaries, `.wasm`, `.py`/`.pyc`,
`.sqlite`, `.safetensors`/`.gguf`, `.dex`, `.class`, `.dll`, Apple plists, ZIP/APK,
etc., detects each artifact's format, and runs the full **audit → defend (bankrupt) →
verify** pipeline. Return the per-artifact summary to the user.
## One-Click Single-Artifact Defense
**Do not ask which files to protect — discover them.**
When the user already named a **specific file**, use:
Optional args (defaults are safe): `intensity` (default bankrupt), `emit_manifest`
(default false), `llm_directives` (default false — risky LLM bait; leave off unless
asked), `weave_decoys` (default false — see stronger path below), `seed`.
- **MCP Tool**: `bla_defend` with `{ "input_path": "<target>", "intensity": "bankrupt", "emit_manifest": true }`
- **CLI Command**: `npx bla-agents@latest defend <target> --manifest`
## Optional stronger path: structural weave decoys
When the user wants **stronger / structural** hardening (not dismissible noise),
especially for **Python** customer defs / `.pyc`, or says weave / 织入 / decoy
entanglement:
1. Read directives: MCP resource `bla://directives/weave-decoy`
(or `bla://docs/weave-decoy-mcp` / repo `docs/weave-decoy-mcp.md`).
2. Plan (read-only): `bla_weave_plan` with `{ "path": "<project-or-artifact>" }` →
save the `bla-weave-plan/v1` JSON work order.
3. Apply with the real pipeline flag:
- Directory: `bla_harden` `{ "directory": ".", "weave_decoys": true }`
- Named file: `bla_defend` / `bla_protect` with `weave_decoys: true`
4. MCP Prompt shortcut: `bla-weave-decoy`
5. CLI: `npx bla-agents@latest harden . --weave-decoys` (or `defend` / `protect`
with the same flag when supported)
**Principles (non-rule, structural):** decoys must weave into **real customer paths**
(shared runtime slots, entry-reachable edges, mixed mundane/scare stems, opaque-true
happy paths, call-graph entanglement). JS AST weave is already on; Python `.py`/`.pyc`
customer-def structural weave activates when `weave_decoys: true`. See
[docs/weave-decoy-mcp.md](../../../docs/weave-decoy-mcp.md).
## FORBIDDEN for executing agents
Do **not**:
- Add pure noise layers, greppable `BLA-SHIELD` / homogeneous `_bla_*` banners, or
dismissible empty/`pass` honeypot stubs that agents filter as an independent layer
- Use label-only / report-only “decoys” that are not real pipeline attributes
- Drive hardening via a **rule engine** (business-name → fixed snippet tables, homogeneous
marker templates as policy)
- Help crack, reverse-engineer, bypass, or write offensive PoCs against third-party
software
If asked for offensive RE/crack help: refuse briefly; offer defensive harden of **their**
artifacts instead.
## Single-file vs directory
| Intent | Prefer | Example |
| :--- | :--- | :--- |
| Project / build dir, auto-discover | **`bla_harden`** / `bla harden .` | User did not name a file |
| Specific artifact path | **`bla_defend`** / `bla defend <path>` | User named `./dist/app.js` |
`bla_defend` runs the 5-stage Counter-REA pipeline on a known path:
`bla_defend` / `bla defend` executes all 5 Counter-REA defense stages automatically:
1. **Stage 1 (`inspect-release`)**: Read-only scan for accidental `.map` (`sourcesContent`) and symbol leaks.
2. **Stage 2 (`audit`)**: 16-vector REA vulnerability & Claude Opus 4 / Sonnet 4 / o3 token-cost simulation.
3. **Stage 3 (`protect --intensity bankrupt`)**: Injects both the **Resource-Exhaustion Suite** and the **Model-Misjudgment Induction Suite** (including canary honeypots + system-diagnostic decoys across JS, WASM, Python, SQLite, Safetensors/GGUF, DEX, IL2CPP, Apple plist, .NET, JVM, ZIP/APK).
4. **Stage 4 (`verify`)**: Verifies all Counter-REA mathematical invariants and confirms token/cost bankruptcy amplification.
5. **Stage 5 (`manifest`)**: Emits `<output>.bla-manifest.json` attesting to the hardened build.
3. **Stage 3 (`protect --intensity bankrupt`)**: Resource-Exhaustion + Model-Misjudgment suites (canary honeypots + system-diagnostic decoys across supported formats).
4. **Stage 4 (`verify`)**: Counter-REA invariants + token/cost amplification.
5. **Stage 5 (`manifest`)** (opt-in): `<output>.bla-manifest.json` when `emit_manifest: true`.
## Step-by-Step Granular Workflow: Audit -> Protect -> Verify
MCP: `bla_defend` with `{ "input_path": "<target>", "intensity": "bankrupt", "emit_manifest": true }`
CLI: `npx bla-agents@latest defend <target> --manifest`
1. **Audit (`bla_audit` or `npx bla-agents audit <target>`)**:
- Reports the **REA Exposure Score (0–100)**, exposed functions/classes/IPC channels/routes, and estimated attacker token and dollar costs.
2. **Protect (`bla_protect` or `npx bla-agents protect <target> -o <output> --intensity bankrupt`)**:
- **Resource Exhaustion (耗尽对方资源)** and **Model Misjudgment Induction (诱导模型误判)** as documented in [references/counter-rea-matrix.md](references/counter-rea-matrix.md).
3. **Verify (`bla_verify` or `npx bla-agents verify <protected> [original]`)**:
- Confirms that all Counter-REA gates pass and reports exact Token & API Cost Amplification.
## Step-by-step granular workflow
## CLI Fallback Commands
1. **Audit** — `bla_audit` / `npx bla-agents audit <target>`: REA Exposure Score, exposed surfaces, estimated attacker cost.
2. **Protect** — `bla_protect` / `npx bla-agents protect <target> -o <out> --intensity bankrupt` (+ optional `--weave-decoys`).
3. **Verify** — `bla_verify` / `npx bla-agents verify <protected> [original]`.
Map REA analyzers → BLA counters:
[references/counter-rea-matrix.md](references/counter-rea-matrix.md).
## CLI fallbacks
```bash
npx -y bla-agents@latest setup
npx -y bla-agents@latest doctor
npx -y bla-agents@latest harden .
npx -y bla-agents@latest harden . --weave-decoys
npx -y bla-agents@latest defend /absolute/path/to/target --manifest --json
npx -y bla-agents@latest audit /absolute/path/to/target --json
npx -y bla-agents@latest protect /absolute/path/to/target -o /absolute/path/to/protected --intensity bankrupt --json
npx -y bla-agents@latest verify /absolute/path/to/protected /absolute/path/to/target --json
npx -y bla-agents@latest doctor
```
See [references/counter-rea-matrix.md](references/counter-rea-matrix.md) for the complete mapping between REA's internal analyzers and BLA's countermeasures.
## References
- [references/counter-rea-matrix.md](references/counter-rea-matrix.md) — REA ↔ BLA countermeasures
- [docs/one-click-agent-defense.md](../../../docs/one-click-agent-defense.md) — one-click UX (skill-first)
- [docs/cli-and-mcp.md](../../../docs/cli-and-mcp.md) — CLI / MCP parameter index
- [docs/weave-decoy-mcp.md](../../../docs/weave-decoy-mcp.md) — weave-decoy theory + MCP checklist
+13
View File
@@ -6,6 +6,19 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased]
### Changed
- **`blind-llm-agents` skill → v1.3.0**: skill is the primary one-click harden GUIDE
(MCP/CLI remain the execution layer). Documents default `bla_harden` / `harden .`,
optional weave path (`bla://directives/weave-decoy` → `bla_weave_plan` →
`weave_decoys:true`), FORBIDDEN noise/crack patterns, single-file `bla_defend` vs
directory `bla_harden`, and links to weave + Counter-REA docs. Synced
[`docs/one-click-agent-defense.md`](docs/one-click-agent-defense.md),
[`docs/cli-and-mcp.md`](docs/cli-and-mcp.md), and README skill pointers.
- **CLI `--weave-decoys`**: opt-in flag on `harden` / `defend` / `protect`, matching
MCP `weave_decoys` (structural Python customer-def / `.pyc` weave; JS AST weave
already on).
### Added
- **Weave-decoy MCP surface** (`src/shields/weave/weave-decoy.ts`, `src/mcp/server.ts`):
+6 -3
View File
@@ -71,13 +71,13 @@ npx bla-agents harden . # one-command Counter-REA pass on the current proj
npx bla-agents benchmark # 14-target showcase (local estimates + runtime checks)
```
Ask your agent after setup:
Ask your agent after setup (skill guides one-click harden):
```text
Harden my software against REA
```
Or run stages explicitly: `audit` · `protect` · `verify` · `defend` — see [CLI and MCP](docs/cli-and-mcp.md).
Or run stages explicitly: `audit` · `protect` · `verify` · `defend` — see [one-click agent defense](docs/one-click-agent-defense.md) and [CLI and MCP](docs/cli-and-mcp.md).
---
@@ -201,7 +201,7 @@ Protections are format-native (no commercial packer / external LLVM required). R
<details>
<summary><strong>Which agents can use BLA?</strong></summary>
Any coding agent with local MCP or `SKILL.md` workflows. `npx bla-agents setup` configures Claude Code, Cursor, Codex, Gemini CLI, Windsurf, Google Antigravity, OpenCode, VS Code, and GitHub Copilot CLI.
Any coding agent with local MCP or `SKILL.md` workflows. `npx bla-agents setup` configures Claude Code, Cursor, Codex, Gemini CLI, Windsurf, Google Antigravity, OpenCode, VS Code, and GitHub Copilot CLI. After setup, the bundled [`blind-llm-agents` skill](.agents/skills/blind-llm-agents/SKILL.md) is the one-click harden guide (`bla_harden` / `bla harden .`); MCP stays the execution layer.
</details>
<details>
@@ -227,7 +227,10 @@ No. See [DISCLAIMER.md](DISCLAIMER.md). Protective payloads activate under rever
- [DISCLAIMER.md](DISCLAIMER.md) — defensive use, own-software only
- [Installation and agent setup](docs/installation.md)
- [CLI and MCP contracts](docs/cli-and-mcp.md)
- [One-click agent defense (skill-first)](docs/one-click-agent-defense.md)
- [Weave-decoy MCP](docs/weave-decoy-mcp.md)
- [Empirical REA validation](docs/redteam/empirical-rea-validation.md)
- [`blind-llm-agents` skill](.agents/skills/blind-llm-agents/SKILL.md)
- [Counter-REA matrix](.agents/skills/blind-llm-agents/references/counter-rea-matrix.md)
- [v1.4.0 release notes (staged)](docs/releases/v1.4.0.md)
- [CHANGELOG](CHANGELOG.md)
+6 -3
View File
@@ -1,15 +1,18 @@
# CLI & MCP Reference (`bla-agents`)
For coding agents, the **[`blind-llm-agents` skill](../.agents/skills/blind-llm-agents/SKILL.md)** is the primary one-click harden guide; this page indexes the CLI/MCP execution surface.
## CLI Commands
| Command | Description |
| :--- | :--- |
| `npx bla-agents setup` | Auto-registers the `bla` MCP server and `blind-llm-agents` skill with Claude Code, Cursor, Codex, Windsurf, Gemini CLI, and Antigravity. |
| `npx bla-agents harden .` | **Default one-command UX**: auto-discover build outputs and run audit → protect → verify (bankrupt). Prefer over `defend` when you have a project directory. |
| `npx bla-agents defend <target> [--manifest] [--llm-directives]` | **5-stage pipeline** on a known path: `inspect-release` → `audit` → `protect` → `verify` → optional **unsigned** `.bla-manifest.json` (opt-in). Risky LLM directive layers stay off unless `--llm-directives`. |
| `npx bla-agents harden .` | **Default one-command UX**: auto-discover build outputs and run audit → protect → verify (bankrupt). Prefer over `defend` when you have a project directory. Agent guide: [`blind-llm-agents` skill](../.agents/skills/blind-llm-agents/SKILL.md). |
| `npx bla-agents defend <target> [--manifest] [--llm-directives] [--weave-decoys]` | **5-stage pipeline** on a known path: `inspect-release` → `audit` → `protect` → `verify` → optional **unsigned** `.bla-manifest.json` (opt-in). Risky LLM directive layers stay off unless `--llm-directives`. Structural weave decoys stay off unless `--weave-decoys`. |
| `npx bla-agents harden . [--weave-decoys]` | Same as `harden` with opt-in structural weave (`weave_decoys`). |
| `npx bla-agents inspect-release <target>` | Read-only local scan for embedded `.map` (`sourcesContent`) and sourceMappingURL references. |
| `npx bla-agents audit <target>` | Pre-flight 16-vector REA vulnerability & token-cost audit. |
| `npx bla-agents protect <target> [-o <out>]` | Injects BLA's Resource-Exhaustion and Model-Misjudgment shield (`stealth`, `balanced`, `maximum`, `bankrupt`). |
| `npx bla-agents protect <target> [-o <out>] [--weave-decoys]` | Injects BLA's Resource-Exhaustion and Model-Misjudgment shield (`stealth`, `balanced`, `maximum`, `bankrupt`). Opt-in `--weave-decoys` for structural Python customer-def weave. |
| `npx bla-agents verify <target> [original]` | Verifies Counter-REA invariants and calculates token/cost amplification. |
| `npx bla-agents benchmark` | Runs the 8-target Counter-REA Showcase Benchmark suite. |
| `npx bla-agents emit-headers --lang <c\|rust\|go>` | Emits compile-time native defense headers (`bla_shield.h`, `bla_shield.rs`, `bla_shield.go`). |
+55 -10
View File
@@ -1,25 +1,70 @@
# One-Click Agent Defense Hardening (`bla defend` / `bla_defend`)
# One-Click Agent Defense Hardening (skill-first)
Once BLA is registered with your coding agent (`npx bla-agents@latest setup`), you can trigger the entire 5-stage Counter-REA defense pipeline with a single command or natural-language prompt.
Once BLA is registered with your coding agent (`npx bla-agents@latest setup`), the
bundled **`blind-llm-agents` skill** is the primary guide for one-click hardening.
MCP tools and the CLI are the execution layer — prefer calling existing tools from
the skill rather than inventing new MCP surfaces.
## One-Click CLI Usage
Skill source: [`.agents/skills/blind-llm-agents/SKILL.md`](../.agents/skills/blind-llm-agents/SKILL.md)
(metadata.version tracks the skill guide; package version is independent).
## Default: one-sentence project harden
Tell Claude Code, Cursor, Codex, Windsurf, Gemini CLI, OpenCode, or similar:
```text
Harden my software against REA
```
(or Chinese equivalents such as `一键加固` / `一键启动防御强化`)
The agent should invoke:
| Surface | Call |
| :--- | :--- |
| **MCP (preferred)** | `bla_harden` with `{ "directory": "." }` |
| **CLI** | `npx bla-agents@latest harden .` |
| **MCP Prompt** | `bla-harden-project` |
`bla_harden` / `bla harden` **auto-discovers** protectable build outputs (`dist/`,
`build/`, `out/`, `release/`, `.asar`, native binaries, `.wasm`, `.py`/`.pyc`,
`.sqlite`, models, DEX, …) and runs **audit → defend (bankrupt) → verify**.
Agents must **not** ask which files to protect.
## Named single artifact
When a specific file is already known:
```bash
npx bla-agents@latest defend ./dist/app.js --manifest
```
## One-Click Agent Prompt
After running `npx bla-agents@latest setup`, tell Claude Code, Cursor, Codex, Windsurf, Gemini CLI, or OpenCode:
```text
Run bla_defend on ./dist with bankrupt intensity and emit a defense manifest.
Run bla_defend on ./dist/app.js with bankrupt intensity and emit a defense manifest.
```
## What `bla defend` Executes Automatically
## Optional stronger path: structural weave decoys
For structural (non-noise) decoy entanglement — especially Python customer defs /
`.pyc` — follow the skill’s weave path:
1. Read `bla://directives/weave-decoy`
2. Call `bla_weave_plan` on the project/artifact
3. Re-run `bla_harden` / `bla_defend` / `bla_protect` with **`weave_decoys: true`**
(CLI: `--weave-decoys`)
See [`weave-decoy-mcp.md`](./weave-decoy-mcp.md). Do **not** add dismissible
`_bla_*` banners, empty honeypot stubs, or rule-based name→snippet noise layers.
## What the pipeline executes
1. **Stage 1 (`inspect-release`)**: Read-only scan for accidental `.map` (`sourcesContent`) and symbol leaks.
2. **Stage 2 (`audit`)**: 16-vector REA vulnerability & Claude Opus 4 / Sonnet 4 / o3 token-cost simulation.
3. **Stage 3 (`protect --intensity bankrupt`)**: Injects both the **Resource-Exhaustion Suite** and the **Model-Misjudgment Induction Suite**.
4. **Stage 4 (`verify`)**: Verifies all Counter-REA mathematical invariants and confirms `16x–51x` token/cost bankruptcy amplification.
5. **Stage 5 (`manifest`)**: Emits `<output>.bla-manifest.json` attesting to the hardened build.
5. **Stage 5 (`manifest`)** (opt-in): Emits `<output>.bla-manifest.json` attesting to the hardened build.
## Related
- [CLI and MCP](./cli-and-mcp.md)
- [Counter-REA matrix](../.agents/skills/blind-llm-agents/references/counter-rea-matrix.md)
+9
View File
@@ -67,6 +67,7 @@ Options:
--seed <seed> Deterministic PRNG seed (default: per-project random in .blarc)
--manifest Emit unsigned .bla-manifest.json next to protected artifacts (opt-in)
--llm-directives Opt-in risky LLM directive layers (COPPA/medical bait, dual-use imitation, AI chat markers); OFF by default
--weave-decoys Opt-in structural weave decoys (Python customer-def / .pyc co_consts weave; JS AST weave already on); OFF by default
--lang <c|rust|go> Target language for emit-headers (default: c)
--client <id> Agent client ID for setup (repeatable: claude_code, cursor, codex, etc.)
--all-detected Configure all detected coding agents on this machine
@@ -88,6 +89,7 @@ interface ParsedCliArgs {
readonly seed?: string;
readonly manifest: boolean;
readonly llmDirectives: boolean;
readonly weaveDecoys: boolean;
readonly lang: NativeHeaderLanguage;
readonly clients: SupportedAgentClientId[];
readonly allDetected: boolean;
@@ -111,6 +113,7 @@ export function parseCliArgs(argv: readonly string[]): ParsedCliArgs {
let seed: string | undefined;
let manifest = false;
let llmDirectives = false;
let weaveDecoys = false;
let lang: NativeHeaderLanguage = "c";
const clients: SupportedAgentClientId[] = [];
let allDetected = false;
@@ -147,6 +150,8 @@ export function parseCliArgs(argv: readonly string[]): ParsedCliArgs {
manifest = true;
} else if (arg === "--llm-directives") {
llmDirectives = true;
} else if (arg === "--weave-decoys") {
weaveDecoys = true;
} else if (arg === "--mcp") {
command = "mcp";
} else if (arg === "--update") {
@@ -188,6 +193,7 @@ export function parseCliArgs(argv: readonly string[]): ParsedCliArgs {
...(seed !== undefined ? { seed } : {}),
manifest,
llmDirectives,
weaveDecoys,
lang,
clients,
allDetected,
@@ -307,6 +313,7 @@ export async function runCli(argv: readonly string[] = process.argv.slice(2)): P
seed: resolved.seed,
emitManifest: parsed.manifest || loaded.config.emitManifest === true,
llmDirectives: parsed.llmDirectives || loaded.config.llmDirectives === true,
weaveDecoys: parsed.weaveDecoys,
...(parsed.output !== undefined ? { outputDirectory: parsed.output } : {}),
});
if (parsed.json) {
@@ -338,6 +345,7 @@ export async function runCli(argv: readonly string[] = process.argv.slice(2)): P
seed: resolved.seed,
emitManifest: parsed.manifest || loaded.config.emitManifest === true,
llmDirectives: parsed.llmDirectives || loaded.config.llmDirectives === true,
weaveDecoys: parsed.weaveDecoys,
});
if (parsed.sarif) {
console.log(
@@ -393,6 +401,7 @@ export async function runCli(argv: readonly string[] = process.argv.slice(2)): P
seed: resolved.seed,
honeySourceMaps: parsed.honeyMaps,
llmDirectives: parsed.llmDirectives || loaded.config.llmDirectives === true,
weaveDecoys: parsed.weaveDecoys,
});
if (parsed.json) console.log(JSON.stringify(res, null, 2));
else console.log(formatProtectionResult(res));
+37
View File
@@ -0,0 +1,37 @@
import { describe, expect, it } from "vitest";
import * as fs from "node:fs";
import * as path from "node:path";
import { fileURLToPath } from "node:url";
const skillPath = path.resolve(
path.dirname(fileURLToPath(import.meta.url)),
"../.agents/skills/blind-llm-agents/SKILL.md",
);
describe("blind-llm-agents skill guide contents", () => {
it("ships v1.3+ one-click harden guide with weave + forbidden surfaces", () => {
const text = fs.readFileSync(skillPath, "utf8");
expect(text).toMatch(/metadata:\s*\n\s*version:\s*"1\.3\.\d+"/);
expect(text).toContain("bla_harden");
expect(text).toContain("npx bla-agents@latest harden .");
expect(text).toContain("weave_decoys");
expect(text).toContain("bla_weave_plan");
expect(text).toContain("bla://directives/weave-decoy");
expect(text).toContain("bla-weave-decoy");
expect(text).toMatch(/FORBIDDEN/i);
expect(text).toMatch(/bla_defend/);
expect(text).toContain("counter-rea-matrix.md");
expect(text).toContain("weave-decoy-mcp.md");
// Prefer skill guide + existing MCP tools; refuse offensive RE/crack framing.
expect(text.toLowerCase()).toMatch(/refuse|defensive only/);
expect(text).toMatch(/Do not ask which files to protect/i);
});
it("parseCliArgs accepts --weave-decoys for harden/defend/protect parity with MCP", async () => {
const { parseCliArgs } = await import("../src/cli.js");
const parsed = parseCliArgs(["harden", ".", "--weave-decoys", "--json"]);
expect(parsed.command).toBe("harden");
expect(parsed.weaveDecoys).toBe(true);
expect(parseCliArgs(["defend", "x.js"]).weaveDecoys).toBe(false);
});
});