The Agent Skills Directory

[COMMAND_EXECUTION]: Employs standard shell utilities such as ls, find, cat, and the GitHub CLI (gh) to analyze the target repository's language, framework, and directory layout during the discovery phase.
[COMMAND_EXECUTION]: Performs file system operations including directory creation (mkdir), symbolic link management (ln, readlink), and file relocation (mv) to configure the workspace. The logic includes safety checks to prevent overwriting existing user files or directories and verifies tool availability before execution.
[EXTERNAL_DOWNLOADS]: Fetches and installs supplementary skills from the vendor's repository (Fandhe-AI/agent-cli-skills) using the npx skills add command. This represents normal vendor functionality for extending agent capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes local repository files, establishing a data ingestion surface for the agent's configuration design process.
Ingestion points: Reads project manifest files (e.g., package.json, Cargo.toml, pyproject.toml) and GitHub Actions workflow files to determine the build and test environment.
Boundary markers: The instructions do not specify explicit delimiters for the content read from these files, although the data is used for metadata analysis rather than direct execution.
Capability inventory: The agent has access to gh, npx, and standard Unix file management tools throughout the setup workflow.
Sanitization: The skill provides explicit guidance to avoid hardcoding secrets in session hooks and warns against the direct interpolation of user input into shell command hooks.

init-claude