The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides detailed instructions for creating shell scripts that execute AI agent CLI tools such as claude, codex, gemini, copilot, and gh agent-task to perform automated software development tasks.
It facilitates the use of powerful tools such as Bash and Edit within autonomous loops to modify files and execute repository commands.
Best practices like using timeout for execution limits and checking for structured output errors are encouraged to maintain control over automated actions.
[PROMPT_INJECTION]: The skill identifies indirect prompt injection as a risk when agents process untrusted data like log files or pull request diffs.
It instructs the AI to wrap such content in delimited blocks (e.g., <diff>...</diff>) and includes system prompts that tell the model to treat delimited content strictly as data.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and installation paths for AI tools from well-known technology organizations, including Anthropic's Claude Code, Google's Gemini CLI, and GitHub's Copilot infrastructure.
It describes the use of specific GitHub Actions, such as anthropics/setup-claude-code and openai/codex-action, for integrating these agents into CI/CD pipelines.

automate