The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple setup scripts and system commands to configure the local environment. It installs a Chrome Native Messaging Host manifest using a dynamic Node.js execution that specifies broad permissions (allowedOrigins: ['*']), which could allow unauthorized browser components to communicate with the host process.
[CREDENTIALS_UNSAFE]: Manages sensitive API tokens by interacting with system-level credential stores. It writes tokens to the macOS Keychain, Linux secret-tool, and Windows Credential Manager, and modifies shell profile files (~/.zshrc, ~/.bashrc) to export API keys as environment variables, making them accessible to any process in the user's shell session.
[EXTERNAL_DOWNLOADS]: Orchestrates the installation of numerous security tools and repository dependencies through pnpm and specialized installer scripts.
[PROMPT_INJECTION]: The skill ingests data from the local environment (git configuration, tool versions) which acts as an indirect input surface. \n
Ingestion points: Phase 0 Preflight checks in SKILL.md. \n
Boundary markers: Not present. \n
Capability inventory: Full Bash, Edit, and Write access to the filesystem. \n
Sanitization: Validation is limited to version string requirements (e.g., Node >= 22.6).

setup-repo