ismirhas

Mcpaudit — Security skill for Claude Code

Security community

Static security audit of MCP and AI-agent configs (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf).

How to install Mcpaudit

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ismirhas/mcpaudit and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Mcpaudit does

Static security audit of MCP and AI-agent configs (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf). Finds hardcoded secrets, unpinned packages, broad file access, privileged containers and risky tool combinations. OWASP LLM Top 10 + CWE references. Python stdlib only.

Alternatives in Security

Trail Of Bits Claude-code-config — Opinionated production defaults from a top security firm: sandboxing, permissions, hooks, skills, MCP server c 1.6k ★
Pocketpaw — Your AI agent in 30 seconds 775 ★
Agent Skills Guard — 一款提供Agent Skills安全扫描和可视化管理的桌面应用 A desktop application that provides security scanning and visual management fo 389 ★

README

mcpaudit

Static security audit for **MCP and AI-agent configuration files**.

AI agents are only as safe as the tools wired into them. `mcpaudit` reads the config files for Claude Desktop, Claude Code, Cursor, VS Code and Windsurf, and reports what an attacker (or a prompt-injected agent) could do with the access you have granted: leaked secrets, unpinned packages, whole-disk access, privileged containers, auto-approved tool calls and risky capability combinations.

Python standard library only. No dependencies, no network access, no telemetry.
Every finding says what is wrong, why it matters and how to fix it, with an OWASP LLM Top 10 (2025) and CWE reference.
Secret values are never printed. Findings carry only the location and the length.
Output as text, JSON or SARIF 2.1.0 (GitHub code scanning).
Honest about its limits: results are heuristics, and the report states what was and was not scanned.

Quick start

git clone https://github.com/ismirhas/mcpaudit.git && cd mcpaudit
python -m mcpaudit scan demo/vulnerable-mcp.json      # see it find things
python -m mcpaudit scan                               # auto-discover your real configs
python -m mcpaudit scan .mcp.json --fail-on medium    # use in CI
python -m mcpaudit scan . -f sarif -o mcpaudit.sarif  # for GitHub code scanning
python -m mcpaudit rules                              # list all rules

Or install it: `pip install .` gives you the `mcpaudit` command.

Exit codes: `0` clean, `1` findings at or above `--fail-on` (default `high`), `2` usage error or a file that could not be parsed.

What it checks

Rule	Checks for	Severity
MCP001	Hardcoded secrets in `env`, `headers`, args and URLs (vendor key formats and credential-like names)	high / critical
MCP002	Unpinned packages and images (`npx`, `uvx`, `pipx`, `docker run`)	medium / low
MCP003	Whole-disk, home-directory or credential-directory access (`/`,