agentleFS
jackchukajackchuka/skills/claude-permissions-audit/SKILL.md

claude-permissions-audit

Skill15 starsChanged 9 days agoMITView on GitHub

Review and reorganize Claude Code permission settings across all config files (global settings.json, project settings.local.json, dotfiles copies). Identifies redundancy, misplaced permissions, and lack of read/write organization. Produces a clean layout where global settings are the source of truth and project-local files only contain project-specific overrides. Use this skill whenever the user mentions reviewing permissions, cleaning up settings, auditing allowed tools, reorganizing settings.json, or asking "what permissions do I have". Also use when adding new MCP servers or tools and wanting to decide what to pre-allow. Triggers: "review permissions", "audit settings", "clean up settings.json", "permissions audit", "/permissions-audit".

  • Installs packages

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    After adding a new MCP server and deciding which operations to pre-allow

    line 35
  2. Distinctive:

    Read all of them in parallel.

    line 50
  3. Very specific:

    Do NOT use JSON comments (// or /* */) — settings.json must be valid JSON.

    line 74
  4. Distinctive:

    Instead, rely on the ordering of entries to convey grouping.

    line 75
  5. Distinctive:

    Only keep entries that are genuinely project-specific:

    line 93
  6. Distinctive:

    Before applying, ask the user about edge cases.

    line 100
  7. Very specific:

    Scope of CLI tools: Should gh:* (all subcommands) go global, or keep read-only globally with broader access per-project?

    line 102
  8. Distinctive:

    Environment separation: Should dev and prod data sources have the same read access, or should prod be more restrictive?

    line 104
  9. Distinctive:

    Write the updated global settings file

    line 113
  10. Distinctive:

    Write the trimmed project-local settings file

    line 114
  11. Distinctive:

    Update any dotfiles copies to stay in sync

    line 115

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Permissions Audit

Review and reorganize Claude Code permission settings so global settings are the source of truth, project-local files stay minimal, and read vs write operations have clear separation.

When to Use

  • Reviewing what tools and commands are pre-allowed
  • Cleaning up accumulated ad-hoc permission entries
  • After adding a new MCP server and deciding which operations to pre-allow
  • Ensuring global vs project-local settings are properly split
  • Periodic hygiene check on permission sprawl

Phase 1: Discover All Settings Files

Scan for every settings file that contributes to the effective permission set.

Locations to check:

  • ~/.claude/settings.json — global settings (source of truth)
  • <project>/.claude/settings.json — project shared settings (committed)
  • <project>/.claude/settings.local.json — project local settings (gitignored)
  • ~/.claude/.claude/settings.local.json — home directory project local
  • Any dotfiles-managed copy (e.g. home/.claude/settings.json in a dotfiles repo)

Read all of them in parallel. If a file doesn't exist, note it and move on.

Phase 2: Audit

For each permission entry across all files, assess:

  1. Redundancy — Is this entry already covered by a broader rule elsewhere?
    • Example: Bash(gws calendar events:*) is redundant when Bash(gws:*) exists globally
  2. Misplacement — Is this general-purpose but stuck in a project-local file?
    • Example: Bash(python3:*) in project local when it's useful everywhere
  3. Inconsistent granularity — Are similar tools allowed at different specificity levels?
    • Example: Global has gh pr list, gh pr view individually, but project local has gh:*
  4. Missing read/write distinction — Are read-only and write operations mixed together?
  5. Stale entries — One-time approvals that accumulated (e.g., a specific git command with hardcoded paths)

Present findings as a table with issues and proposed resolution for each.

Phase 3: Propose Reorganization

Design the new layout following these principles:

Global settings (source of truth)

Organize permissions into labeled sections by grouping related entries together. Do NOT use JSON comments (// or /* */) — settings.json must be valid JSON. Instead, rely on the ordering of entries to convey grouping. Within the allow array, cluster related permissions together in this order:

Section order (group entries in this sequence, no separators needed):

filesystem reads            — cat, tree, grep, wc, file reads
<lang> inspection           — go doc, cargo check, linters (no side effects)
GitHub CLI reads            — gh pr list/view/diff, gh issue list (no mutations)
external tool reads         — calendar, slack CLI, other read CLIs
web reads                   — WebSearch, WebFetch with domain allowlist
MCP <service> reads         — list, get, describe, search operations
build / test / install      — go build, cargo test, pnpm install (local side effects)
skills                      — Skill(superpowers:*), etc.

Project-local settings (minimal overrides)

Only keep entries that are genuinely project-specific:

  • Broader tool access needed for a specific workflow (e.g., Bash(gh:*) for a repo that creates PRs frequently)
  • Project-specific MCP tools not used elsewhere
  • Temporary elevated access during active development

Clarifying questions

Before applying, ask the user about edge cases. Common questions:

  • Scope of CLI tools: Should gh:* (all subcommands) go global, or keep read-only globally with broader access per-project?
  • MCP write operations: Which write operations (send message, create issue) should stay prompted vs pre-allowed?
  • Environment separation: Should dev and prod data sources have the same read access, or should prod be more restrictive?
  • Dotfiles sync: Does the user maintain a dotfiles copy that needs updating?

Adapt questions to what's actually ambiguous — don't ask about things that are obvious from context.

Phase 4: Apply Changes

Once the user confirms the plan:

  1. Write the updated global settings file
  2. Write the trimmed project-local settings file
  3. Update any dotfiles copies to stay in sync
  4. Clear out entries from other local files that were absorbed into global

Phase 5: Summary

Present a before/after comparison:

FileBeforeAfter
GlobalN flat entriesM entries in K sections
Project localX entriesY entries (project-specific only)

List what moved where, and what still requires prompting (write operations not pre-allowed).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.