Grades are a hygiene check, not a security guarantee.

MCP, MCP server

Mcp Audit

See what your AI agents can access - before they go live.

Compatibility: Not stated.

Grade C

One Medium finding. Read the findings before installing.

Graded by PlugVetRubric v2

Not a guarantee. A hygiene check of the source at the commit shown, not a security audit.

Rescanned at commit 6ae4e1ff58f3, unchanged since the check below.

AGENT_IGNORE_INSTRUCTIONS

Why this grade

  • AGENT_IGNORE_INSTRUCTIONS Medium

    Text tells the agent to ignore earlier instructions.

    docs/MCP-SECURITY-TESTING-STRATEGY.md, line 491 | Prompt injection | Inject via tool inputs | `Ignore previous instructions...` |

Any High, or two or more Medium findings, give D. One Medium gives C. Low findings or a security Note give B. Each code names a finding behind the letter.

If you think this grade is wrong, email [email protected] with the entry link and your evidence. We aim to acknowledge within two working days and decide within seven. If we confirm an error, we aim to correct it within 24 hours of confirming it. These are aims, not guarantees.

Grades are a hygiene check of the source at the commit shown, on the date checked, not a security audit or a guarantee. The free directory grades are automated. Before release, a PlugVet engineer reviews every paid report and any grade we hold back, including all F grades, against the source code and the evidence we quote.

What was checked

Plain notes from the hygiene checks.

  • Permissions

    Note

    requests shell access. The manifest names this permission.

    PERMISSION_SHELL_DECLARED

    mcp_audit/models.py, line 23

    api_used: str # e.g., "child_process.exec", "subprocess.run(shell=True)"

    mcp_audit/commands/scan.py, line 504

    "child_process.exec / subprocess.run(shell=True). See "

    mcp_audit/data/allowlist_bypass_binaries.py, line 18

    "python": "-c '<code>' executes arbitrary Python, including os.system()/subprocess calls",

    Show more

    mcp_audit/data/allowlist_bypass_binaries.py, line 19

    "python3": "-c '<code>' executes arbitrary Python, including os.system()/subprocess calls",

    mcp_audit/data/owasp_llm.py, line 29

    eval calls without sanitization - e.g., child_process.exec(`gh ${args.repo}`) is exploitable becau

    mcp_audit/data/risk_definitions.py, line 166

    exec / execSync / util.promisify(exec), subprocess.run(shell=True), os.system) with arguments

    mcp_audit/data/risk_definitions.py, line 167

    ll=True with user-controlled input; use subprocess.run([cmd, arg1, arg2]) with a list. Regex /

    mcp_audit/data/risk_definitions.py, line 172

    "In Python, drop `shell=True` and pass `subprocess.run([cmd, *args])` with a list.",

  • Install and update

    Pass

    No install-time script found, and dependencies are pinned or absent.

  • Network

    Concern

    can reach a host that is not written as a fixed address. can reach api.github.com. The description does not name that host.

    NETWORK_NO_FIXED_HOST NETWORK_UNNAMED_HOST

    app.js, line 17

    fetch(`${ANALYTICS_URL}?${params}`, {

    app.js, line 1125

    const response = await fetch('demo-data.json');

    app.js, line 1785

    const response = await fetch(`https://api.github.com${endpoint}`, {

    Show more

    app.js, line 3203

    const response = await fetch(REPORT_API_URL, {

    backend/api/report.ts, line 15

    await fetch(`${ANALYTICS_URL}?${params.toString()}`, {

    mcp_audit/commands/scan.py, line 976

    response = requests.post(

    mcp_audit/commands/trust.py, line 184

    with urllib.request.urlopen(req, timeout=10) as response:

    mcp_audit/commands/trust.py, line 214

    with urllib.request.urlopen(req, timeout=10) as response:

  • Secrets

    Concern

    reads a credential or environment file.

    SECRETS_ENV_READ SECRETS_PATH_DOTENV

    app.js, line 992

    const env = config?.env || {};

    app.js, line 2243

    const env = r.rawConfig?.env || {};

    backend/api/report.ts, line 75

    const expectedKey = (process.env.MCP_AUDIT_API_KEY || '').trim();

    Show more

    backend/lib/email.ts, line 8

    pass: process.env.GMAIL_APP_PASSWORD // 16-char app password

    mcp_audit/scanners/project.py, line 3

    Scans directories for MCP config files, .env files, and MCP dependencies

    mcp_audit/scanners/project.py, line 29

    ".env",

    mcp_audit/scanners/project.py, line 30

    ".env.local",

    mcp_audit/scanners/project.py, line 31

    ".env.development",

    mcp_audit/scanners/project.py, line 32

    ".env.production",

    mcp_audit/scanners/project.py, line 69

    # Scan for .env files

  • Agent instructions

    Serious

    instructions tell the agent to ignore earlier instructions.

    AGENT_IGNORE_INSTRUCTIONS

    docs/MCP-SECURITY-TESTING-STRATEGY.md, line 491

    | Prompt injection | Inject via tool inputs | `Ignore previous instructions...` |

  • Obfuscation

    Pass

    No encoded or minified payload found.

  • Licence

    Pass

    A licence file is present (LICENSE).

  • Maintainer activity

    Pass

    The latest commit is within the last 365 days.

Get the full report on this tool, £49

Note

Hygiene check of commit 6ae4e1ff58f3. requests shell access. The manifest names this permission. can reach a host that is not written as a fixed address. can reach api.github.com. The description does not name that host. reads a credential or environment file. instructions tell the agent to ignore earlier instructions.

Date and source

Date checked:

View source

github.com

Commit graded: 6ae4e1ff58f3a75ca356fd173112770d217f59a3

This link goes to the original host. PlugVet does not host or check what you download from there, and installing it is your own decision and risk.

Read how this letter is chosen