claude-settings-auditor
Skill1 starsChanged 5 days agoMITView on GitHub
Audit every Claude Code settings layer — user, project, local, and managed — and report the effective merged configuration with its risks: over-broad Bash allows, missing deny rules for secrets, bypassPermissions defaults, unvetted MCP servers and hooks, and rules that never match. Use before trusting a new repo's checked-in settings, or to harden your own before handing the agent more autonomy.
- Reads credentials
- Deletes or force-pushes
- Commits and pushes
In this repository163 other files it gives its agents.
- agents.mdAGENTS.md
- 3d-asset-handoff-sheetSkill
- accessibility-regression-auditorSkill
- adr-writerSkill
- agent-memory-designerSkill
- agent-trajectory-evaluatorSkill
- alerting-rules-tunerSkill
- analysis-memo-writerSkill
- annotation-disagreement-ledgerSkill
- annotation-task-spec-builderSkill
- api-deprecation-plannerSkill
- api-error-contract-designerSkill
- architecture-diagram-generatorSkill
- asset-metadata-review-ledgerSkill
- auth-flow-reviewerSkill
- background-job-reliability-auditorSkill
- branch-rebaserSkill
- brand-voice-profilerSkill
- bundle-analyzerSkill
- cache-policy-designerSkill
- canary-release-plannerSkill
- caption-correction-ledgerSkill
- changelog-from-prsSkill
- chart-chooserSkill
- chunking-strategy-optimizerSkill
- circular-dependency-breakerSkill
- code-example-verifierSkill
- cold-email-deliverability-auditorSkill
- cold-start-optimizerSkill
- commit-splitterSkill
- competitor-teardownSkill
- component-spec-writerSkill
- connection-pool-tunerSkill
- content-repurposerSkill
- contract-obligation-registerSkill
- contract-test-designerSkill
- conventional-commitsSkill
- cors-configuratorSkill
- coverage-gap-finderSkill
- crm-export-auditorSkill
- dashboard-designerSkill
- data-retention-auditorSkill
- dataset-first-lookSkill
- dead-code-finderSkill
- deadlock-diagnoserSkill
- decision-log-reconcilerSkill
- dependency-auditSkill
- dependency-upgrade-plannerSkill
- design-brief-writerSkill
- design-critique-checklistSkill
- design-token-extractorSkill
- devcontainer-designerSkill
- distributed-tracing-instrumenterSkill
- dockerfile-optimizerSkill
- email-sequence-drafterSkill
- embedding-index-tunerSkill
- embedding-set-inspectorSkill
- entity-match-candidate-ledgerSkill
- exit-ticket-builderSkill
- extract-moduleSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Collect every layer: Read ~/.claude/settings.json, .claude/settings.json, .claude/settings.local.json, and check for managed policy files …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Claude Code's behavior is the merge of up to five settings sources — and the merged result is what nobody ever reads. This skill reads it for you: every layer, precedence applied, with each risky or dead rule called out and a safer replacement proposed. Run it on a freshly cloned repo before working in it, or on your own setup before granting more autonomy.
When to use this skill
- You cloned a repo with a checked-in
.claude/settings.json(and maybe.mcp.jsonand hooks) and want to know what you're trusting before the first session. - Your permission prompts feel wrong — too many, too few — and you want the effective rule set explained.
- You're about to loosen the reins (acceptEdits, broader allows, CI automation) and want the floor checked first.
When NOT to use this skill
- You want to write a specific hook — that's hook-writer.
- You're auditing the application's code for vulnerabilities — that's the security-auditor agent; this skill audits the agent harness configuration itself.
Instructions
- Collect every layer. Read
~/.claude/settings.json,.claude/settings.json,.claude/settings.local.json, and check for managed policy files (platform-specific admin paths). Include.mcp.jsonand any.claude/hooks/scripts referenced. Note which files exist, which are committed to the repo, and which are personal. - Compute the effective configuration. Apply precedence (managed > local > project > user) and the permission decision order (deny → ask → allow). Produce the merged permissions table, the active hooks by event, the MCP servers by scope, and the notable toggles (
defaultMode,enableAllProjectMcpServers,disableAllHooks,autoMemoryEnabled). - Hunt permission holes. Flag, with severity: blanket allows (
Bash,Bash(*),mcp__*in allow), allows that swallow more than intended (Bash(git *)coversgit push --force), missing deny rules for secrets (.env*, key files, cloud credential paths),WebFetchunbounded, andbypassPermissionsas a default mode anywhere outside a container. - Vet hooks and MCP servers as supply chain. For each hook script: what does it execute, is the source readable, does any blocking hook fail open? For each MCP server: scope, provenance, whether secrets are inline instead of
${VAR}expansion, and whetherenableAllProjectMcpServersauto-approves more than the user realizes. Read the scripts — don't trust filenames. - Find dead and shadowed rules. Rules that can never fire (shadowed by an earlier deny, malformed pattern, the
Bash(ls *)-vs-lsofword-boundary trap, tools that don't exist) are noise that breeds false confidence — list them for deletion. - Report by severity, then propose the patch. CRITICAL (acts now, dangerous), WARN (risky under the wrong prompt), INFO (hygiene). For each finding: the file and line, why it matters in one sentence, and the exact replacement rule. Offer the corrected settings JSON as a diff the user can apply — but do not modify files unless asked.
[!NOTE] A checked-in settings file is the team's contract — recommend fixes to it as a PR suggestion, not a silent local edit, and keep personal loosening in
settings.local.jsonwhere it belongs.
[!TIP] The fastest wins are nearly universal: add
deny: ["Read(./.env)", "Read(./.env.*)", "Read(./secrets/**)"], replace any bareBashallow with the three or four commands actually used, and movegit pushtoask.
Output
An effective-configuration summary (what wins, from which file), a severity-ordered findings list with file:line and one-line impact each, and a ready-to-apply corrected settings diff — plus the explicit list of things that looked unusual but are fine, so the next auditor doesn't re-litigate them.
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
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.

