{
  "markdown": "# agent-guard-mcp\n\n[![npm](https://img.shields.io/npm/v/@liminallablibs/agent-guard-mcp)](https://www.npmjs.com/package/@liminallablibs/agent-guard-mcp)\n[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE)\n\n> **One line:** Verify-before-act safety for AI agents — check packages, lockfiles, skill manifests, and CI workflows *before* install or merge. **Zero API keys. Zero config.**\n\n**Scores Cursor/Claude skill manifests + GitHub Actions workflows** — not just npm dependencies.\n\n| When you… | Call |\n|-----------|------|\n| `npm install foo` / agent suggests a new dep | `check_package` |\n| Review an LLM-generated lockfile | `verify_lockfile` |\n| Install a Cursor skill or MCP plugin | `score_manifest` |\n| Merge a PR touching `.github/workflows/*` | `check_workflow` |\n\nCall these *before* installing a dependency, merging a CI change, or installing a third-party skill/plugin — so an agent refuses hallucinated packages, typosquats, poisoned manifests, and compromised CI actions.\n\nExposes four MCP tools (also available as a plain HTTP service):\n\n| Tool | Use before… | Returns |\n|------|-------------|---------|\n| `check_package` | adding a single npm/PyPI dependency | OK / SUSPICIOUS / DANGER + typosquat/slop flags |\n| `verify_lockfile` | running `npm/pip/yarn install` | scans every direct + transitive dep in a lockfile |\n| `score_manifest` | installing a Cursor/Claude skill or MCP/Smithery plugin | poison-signature + scope-overreach + drift score 0–100 |\n| `check_workflow` | merging a PR that touches CI | flags mutable action pins, compromised actions, curl\\|bash, pwn-requests, secret exposure |\n\n## Why\n\nAI agents routinely hallucinate package names (slopsquatting), trust LLM-generated lockfiles, install unvetted skill packs, and pin CI actions to mutable tags. Each is a live supply-chain vector. This server gives an agent a cheap \"is this safe?\" call before it acts.\n\n## Install (MCP client)\n\n```json\n{\n  \"mcpServers\": {\n    \"agent-guard\": { \"command\": \"npx\", \"args\": [\"-y\", \"@liminallablibs/agent-guard-mcp@0.2.1\"] }\n  }\n}\n```\n\n## Agent rule (paste into Cursor Rules / CLAUDE.md)\n\n> Before installing any npm/PyPI package, running install from a lockfile, installing a third-party skill/MCP plugin, or merging CI workflow changes, call the agent-guard MCP tools (`check_package`, `verify_lockfile`, `score_manifest`, `check_workflow`). If verdict is SUSPICIOUS, REVIEW, or BLOCK/DANGER, stop and show findings to the user — do not proceed.\n\n## Quick demo\n\n```bash\nnpx -y @liminallablibs/agent-guard-mcp@0.2.1   # MCP clients spawn this via npx\n# HTTP mirror:\nnpm run http && curl \"http://localhost:8402/check?name=reactt&ecosystem=npm\"\n# → DANGER — typosquat of \"react\"\n```\n\nOr run directly:\n\n```bash\nnpm install\nnode src/mcp-server.mjs        # stdio MCP server\nnpm run http                   # optional HTTP mirror on :8402\n```\n\n## HTTP endpoints (mirror of the MCP tools)\n\n- `GET  /check?name=<pkg>&ecosystem=<npm|pypi>`\n- `POST /check-lockfile   {lockfile_content, format}`   — format ∈ package-lock.json | yarn.lock | pnpm-lock.yaml | poetry.lock | requirements.txt\n- `POST /score-manifest   {manifest_type, manifest_content}`\n- `POST /check-workflow   {workflow_content}`\n\n## License\n\nMIT.\n",
  "bytes": 3248,
  "sha": "90ec64fa22cc5cf358f0c82a7d341bdf49090e140f11e473f7cbc0ad0562835a",
  "repo_slug": "liminalpepe/agent-guard-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_liminalpepe_agent_guard_mcp_a8d4a875/readme"
}