claude-settings-audit

Generate recommended Claude Code settings.json permissions by detecting a repository's tech stack.

1|Updated Apr 16, 2026
One-click install
npx skills add https://github.com/klh/skills --skill claude-settings-audit-klh
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: claude-settings-audit
Source: https://github.com/klh/skills/tree/main/.well-known/agent-skills/settings-audit
Command: npx skills add https://github.com/klh/skills --skill claude-settings-audit-klh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Configuring Claude Code permissions manually is error-prone: teams either allow too much (security risk) or too little (constant approval prompts). This Skill analyzes a repository and produces a tailored settings.json allow list of read-only commands matching the project's actual toolchain. ## Core Features & Use Cases - Tech Stack Detection: Scans for indicator files (package.json, pyproject.toml, go.mod, Cargo.toml, lock files) to identify languages, package managers, frameworks, and monorepo structure. - Permission Generation: Builds a baseline of safe read-only commands (git, gh, ls, cat) plus stack-specific commands, including only the package manager actually used. - MCP & WebFetch Suggestions: Recommends Sentry or Linear MCP server configs for .mcp.json and framework documentation domains for WebFetch. - Use Case: When onboarding Claude Code to a Django project using Poetry and Sentry, run the audit to get a ready-to-paste settings.json with poetry, Django docs domains, and Sentry MCP configuration. ## Quick Start Analyze this repository and generate a recommended Claude Code settings.json with read-only permissions for the detected tech stack.

Frequently Asked Questions about claude-settings-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a Claude Code settings.json for my project?▼

Run the settings audit against your repository root. It detects your tech stack from files like package.json, pyproject.toml, and lock files, then outputs a complete settings.json with read-only bash permissions grouped by category.

What commands should I allow in Claude Code permissions?▼

Allow only read-only commands that cannot modify state, such as ls, cat, git status, git log, and gh pr view. Include stack-specific commands only for tools actually detected in the project, and never include install, build, or delete commands.

Does the audit support monorepos and multiple languages?▼

Yes, it detects monorepo indicators like lerna.json, nx.json, turbo.json, and pnpm-workspace.yaml, and can combine permissions for multiple detected languages such as Python, Node.js, Go, and Rust in one settings.json.

How does it handle multiple package managers like pnpm and npm?▼

It includes commands only for the package manager identified by the lock file present, such as pnpm-lock.yaml or yarn.lock. If pnpm is detected, npm and yarn commands are excluded to keep the allow list minimal.

Can I configure Sentry or Linear MCP servers with this audit?▼

Yes, when Sentry SDK or Linear usage is detected, the audit suggests MCP server entries for .mcp.json, including the Sentry HTTP endpoint and the Linear npx-based server. MCP config goes in .mcp.json, not settings.json.