# Audit Skill Permissions

Claude Code skills can need shell permissions that are hard to review manually. This skill analyzes declared command usage and helps create matching allowlist entries with warnings for broad changes.

## Install

```bash
npx skillstore add guo-yu/skill-permissions
```

## Metadata

- Status: approved
- Slug: guo-yu-skill-permissions
- Skillstore revision: r1
- Version status: missing
- Tree hash: 2ce1786f4143fb8fe46ca22383e8c2d190eb221cf3b3af5075647274a1771c71
- Author: guo-yu
- GitHub username: guo-yu
- License: MIT
- Repository: https://github.com/guo-yu/skills/tree/master/skill-permissions
- Ref: 1ffa7643651792ccb4bd3b15d924d2c97edff755
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, filesystem
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/guo-yu-skill-permissions
- Manifest: https://skillstore.pages.dev/api/skills/guo-yu-skill-permissions/manifest

## Capabilities

- Scans skill documentation for explicit Bash command patterns.
- Maps detected commands to Claude Code permissions.allow rules.
- Shows summaries for one skill or all installed skills.
- Can merge generated allow rules into ~/.claude/settings.json.
- Can add an allow command section to multiple skill files.

## Use Cases

- Review a new community skill: Analyze a single downloaded skill before granting Claude Code Bash permissions.
- Prepare a local skill set: Summarize permissions across installed skills before a team setup session.
- Document permission needs: Generate allow command guidance that explains the permissions a skill requires.

## Prompt Templates

### Analyze one skill

```
Use skill-permissions to analyze the permissions required by <skill-name> and explain each requested permission.
```

### Review before authorization

```
Run the read-only skill permission analysis for <skill-name>. Do not modify settings until I approve the proposed allow rules.
```

### Audit all installed skills

```
Analyze all installed Claude Code skills and group the results by low, medium, and high permission risk.
```

### Plan controlled authorization

```
Analyze all skills, identify permissions that should not be auto-approved, and prepare explicit allow commands for only the reviewed skills.
```

## Limitations

- It relies on documented command patterns and may miss hidden behavior outside SKILL.md.
- Batch modes can grant broad permissions if used without manual review.
- Permission changes require a new session or /clear to take effect.
- It does not prove that a skill is safe to run.

## Best Practices

- Run read-only analysis before using any authorization command.
- Prefer explicit per-skill authorization over allow-all or inject modes.
- Review generated wildcard permissions before adding them to settings.json.

## Anti Patterns

- Do not run batch authorization for unreviewed community skills.
- Do not treat generated allow rules as proof that a skill is safe.
- Do not inject allow commands into many skill files without version control or backups.

## Security Audit

- Audited at: 2026-07-09T04:34:34.718\+00:00
- Summary: The skill is a security-sensitive permission management helper, not malware. Most dangerous command hits are documented blocklist examples, but the skill can broaden Claude Code Bash allow rules and batch-modify settings or skill files, so publication should require strong warnings and safer defaults.

## Stats

- Views: 275
- Downloads: 8
- Favorites: 0
- Popularity score: 0