Dot Claude

dot-claude

When to Use

• Configuring allowed tools and bash permissions for Claude Code.
• Defining PreToolUse or PostToolUse hooks for specific commands or file edits.
• Setting up environment variables or enabling LSP plugins for Claude Code.
• Analyzing or updating the .claude/settings.json workspace configuration.

When Not to Use

• When working with the Anthropic Claude API or Claude documentation.
• When configuring generic coding standard rules instead of Claude-specific behavior.

Core Process

1. Locate or Create Configuration: Ensure the .claude directory exists at the root of the project.
2. Define Permissions: Use .claude/settings.json to explicitly allow specific Bash commands to prevent Claude Code from blocking them.
3. Configure Hooks: Set up PreToolUse or PostToolUse hooks in settings.json to enforce safety checks, trigger auto-formatting, or log activity.
4. Set Environment & Plugins: Enable required tools like language servers (pyright-lsp) or specify environment variables.

Best Practices

• Explicit Permissions: Instead of granting broad permissions, use precise glob patterns (e.g., Bash(uv tool *)).
• Hook Activity Logging: Redirect output from hooks into a specific log file like .claude/hook-activity.log to trace tool side effects.
• Fail-Safe Hooks: Ensure PreToolUse hooks return appropriate exit codes (e.g., exit 2) to block destructive commands.

Common Pitfalls

• JSON Syntax Errors: settings.json must be strictly valid JSON, without trailing commas or comments.
• Overly Broad Matchers: Be precise with matcher regex in hooks to avoid triggering scripts unnecessarily (e.g., Edit|Write vs just Bash).

References

• [Example settings.json](references/vojay-dev/settings.json)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Cogni-AI-OU
• Source: Cogni-AI-OU/cogni-ai-agent-skills
• License: MIT

Install and usage instructions live in the source repository linked above.