MCP.Market

Agent Guard MCP server

by liminalpepe·io.github.liminalpepe/agent-guard-mcp·v0.2.1
No reviews yet

Verify-before-act safety checks for AI agents: packages, lockfiles, manifests, CI workflows.

B80/100grade B
Write a reviewInstall
What users say
No reviews yet
Be the first
Safety scan
B80/100

full report

Adoption
New

0 stars37 downloads/wk

Reviews

Write one

Nobody has reviewed Agent Guard yet.

If you have run it, two minutes of your experience saves the next person an afternoon.

Agent Guard tools (4)

write = sends, deletes, buys or posts

Read from the package source without running it. The installed server may list more.

  • check_package

    Pre-install supply-chain gate for a single npm or PyPI package. Call BEFORE running npm install, pip install, or adding a dependency to package.json. Returns OK | SUSPICIOUS | DANGER (DANGER ~= NO-GO) with typosquat/slopsquat/hallucination flags. Read-only: queries public registries only; does not install anything. Do NOT use for full lockfile audits (use verify_lockfile) or CVE-only scans (use np

  • check_workflow

    Validate CI workflow YAML (GitHub Actions / GitLab CI) BEFORE merging a PR that touches .github/workflows or GitLab CI. Call when reviewing or authoring CI pipelines. Flags mutable action pins (@v1), known-compromised actions, untrusted owners, curl|bash fetch-exec, pull_request_target pwn-request patterns, and secret exposure. Returns risk 0-100 and PROCEED | REVIEW | BLOCK. Read-only. Do NOT use

  • score_manifest

    Score a third-party AI-agent extension manifest (skill/plugin manifest audit) BEFORE install. Call BEFORE installing a Cursor skill, Claude skill, MCP server package, or Smithery plugin. Detects poison/backdoor signatures, credential scope over-reach, and drift vs an approved baseline. Returns risk 0-100 and PROCEED | REVIEW | BLOCK. Supports manifest_type: cursor-skill | claude-skill | mcp-server

  • verify_lockfile

    Scan an entire lockfile (direct + transitive deps) BEFORE npm/pip/yarn/pnpm install. Call when the user pastes or generates a lockfile, or before trusting an LLM-written package-lock.json / yarn.lock / pnpm-lock.yaml / poetry.lock / requirements.txt. Returns per-package OK | SUSPICIOUS | DANGER counts and flagged names. Read-only; pass raw file text, not a filesystem path. Do NOT use for a single

Public scan report

scanner v0.1.9 · 2026-09-26 · same rubric, same numbers if you re-run it

1 medium
  • Code scan8 source files scanned20/25
  • –Live reliabilityno gateway calls yet and no remote to proben/a
  • –Tool poisoningtools not inspected (local package is not executed); not countedn/a
  • Auth qualitylocal package, no credentials required12/15
  • Maintenancelast push 86 days ago12/15
  • Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Findings (1)

  • mediumeval / new Function usedexec.eval
    src/manifest.mjs: … [/eval\s*\(/, 'HIGH', 28, 'dynamic eval()'], [/new\s+Function\s*\(/, 'HIGH',…
Overall 80/100. Components that don't apply are left out of the denominator. Any critical finding is an F.RubricAppeal a findingJSON

How Agent Guard compares

Against every graded Developer tools MCP server, from our own scans and uptime checks. Updated after each scan run.

Safety score
80/100
higher than 55% of 1,119 Developer tools servers
Tools
4
the typical Developer tools server has 7
Tools that change things
0%
typical: 5%. Fewer write tools means less an agent can break
Downloads a week
37
more than 23% of 436 Developer tools servers with a package

What its tools let an agent do: 4 read

Install Agent Guard in Claude Code, Cursor or VS Code

Runs npx -y @liminallablibs/agent-guard-mcp on your machine. Read the scan report first; the gateway never runs local packages.

claude mcp add agent-guard-mcp -- npx -y @liminallablibs/agent-guard-mcp
Add to Cursor

What the publisher says

From the Agent Guard repository's README, as published. We do not edit it. Read it on GitHub

agent-guard-mcp

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.

Scores Cursor/Claude skill manifests + GitHub Actions workflows — not just npm dependencies.

Call 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.

Exposes four MCP tools (also available as a plain HTTP service):

Why

AI 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.

Install (MCP client)

{
  "mcpServers": {
    "agent-guard": { "command": "npx", "args": ["-y", "@liminallablibs/agent-guard-mcp@0.2.1"] }
  }
}

Agent rule (paste into Cursor Rules / CLAUDE.md)

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 (checkpackage, verifylockfile, scoremanifest, checkworkflow). If verdict is SUSPICIOUS, REVIEW, or BLOCK/DANGER, stop and show findings to the user — do not proceed.

Quick demo

npx -y @liminallablibs/agent-guard-mcp@0.2.1   # MCP clients spawn this via npx
# HTTP mirror:
npm run http && curl "http://localhost:8402/check?name=reactt&ecosystem=npm"
# → DANGER — typosquat of "react"

Or run directly:

npm install
node src/mcp-server.mjs        # stdio MCP server
npm run http                   # optional HTTP mirror on :8402

HTTP endpoints (mirror of the MCP tools)

  • GET /check?name=&ecosystem=
  • POST /check-lockfile {lockfile_content, format} — format ∈ package-lock.json | yarn.lock | pnpm-lock.yaml | poetry.lock | requirements.txt
  • POST /score-manifest {manifesttype, manifestcontent}
  • POST /check-workflow {workflow_content}

License

MIT.

Nothing above is checked by us. What we check is on the safety report.

Agent Guard: common questions

Is Agent Guard MCP server safe?
Mostly: it is graded B (80/100). Read the Agent Guard safety report
How do I install Agent Guard?
It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
Does Agent Guard need an API key?
Not as far as the registry entry and our scan can tell: no credentials are declared or required.
Is Agent Guard maintained?
The last commit was 125 days ago (2026-07-02). The latest release is v0.2.1.
What can I use instead of Agent Guard?
Servers from other publishers that do the same job: argot MCP server, zanadir MCP server and Avalanche MCP server.
How does Agent Guard compare with other Developer tools MCP servers?
safety score 80/100 (higher than 55% of 1,119 Developer tools servers); tools 4 (the typical Developer tools server has 7); tools that change things 0% (typical: 5%. Fewer write tools means less an agent can break). Full comparison.

Alternatives to Agent Guard

Same job from other publishers: the closest match first, then the best rated.

  • argot
    Read-only context and complete checks against patterns learned from git history. Local; no LLM.
    B
  • zanadir
    Find what a repository's CI is missing, and which security checks cannot fail the build.
    B
  • Avalanche
    Act on Avalanche from your agent: scaffold a dapp, deploy a contract, mint, and read chain state.
    A
  • Gnt
    Git-native policy layer for AI agents: check_action verdicts against rules approved via PR.
    A

More from liminalpepe →