Permissions Audit — Security skill for Claude Code
Review settings.json permissions — flag overly broad rules, identify likely-unused ones, suggest pruning.
How to install Permissions Audit
Installs to ~/.claude/skills/marianasmall-claude-code-starter-kit-permissions-audit/SKILL.md
mkdir -p ~/.claude/skills/marianasmall-claude-code-starter-kit-permissions-audit && curl -fsSL https://raw.githubusercontent.com/marianasmall/claude-code-starter-kit/HEAD/commands/permissions-audit.md -o ~/.claude/skills/marianasmall-claude-code-starter-kit-permissions-audit/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Permissions Audit does
name: permissions-audit description: Review settings.json permissions — flag overly broad rules, identify likely-unused ones, suggest pruning.
/permissions-audit
Audit `~/.claude/settings.json` permissions for security and tidiness.
What to do
1. Read the settings
Read `~/.claude/settings.json` and parse the `permissions.allow` array.
2. Categorize entries
Group into:
- Wildcard rules (e.g.,
Bash(*),mcp__*__*) — broadest, highest risk - **Tool-prefix wildcard
Alternatives in Security
- Security Ownership Map — Map people-to-file ownership, compute bus factor, and identify risks 14.6k ★
- Trail Of Bits Claude-code-config — Opinionated production defaults from a top security firm: sandboxing, permissions, hooks, skills, MCP server c 1.6k ★
- Reflexion — Audit and optimize the .claude/ ecosystem — skills, hooks, guides, patterns, agents, and settings — for qualit 224 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Audit Policies
Analyze a project's Claude Code permission settings and suggest toolgate policies. Use when the user asks to a
J Config Audit
Security audit of Claude, Codex, Gemini, and shared agent configuration for secrets, over-broad permissions, a
Security Actions
GitHub Actions for AI agent and supply chain security checks: audit agent configs (Claude Code settings, MCP s
Configuration Review
Analyze configuration management and environment settings to identify security risks and improvement opportuni
Harden Deps
Deployment hardening step 3 — run ecosystem-native vulnerability audit (full tree AND production-only, so ship
Agentconfig Doctor
MCP security scanner & config linter for AI coding agents. Finds leaked secrets, risky commands, broad permiss
Related Agents
Cr Silent Failure Hunter
Hunts silent failures in the PR diff — empty or overly broad catch blocks, fallbacks that mask errors, excepti
Claude Code Infra
Use PROACTIVELY and automatically — do not wait to be asked — to configure Claude Code itself: sub-agents, ski
C2 Detection
Flag likely command-and-control / beaconing patterns.