The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides extensive documentation and examples for automating development workflows through shell scripts triggered by lifecycle hooks (PreToolUse, PostToolUse, SessionStart, etc.). These scripts facilitate automated formatting, linting, testing, and validation of code changes.
[DYNAMIC_EXECUTION]: The skill documents the integration of Model Context Protocol (MCP) servers using various runtimes including Node.js, Python, and Docker. These integrations allow the agent to invoke external tools and services via standard interfaces.
[INDIRECT_PROMPT_INJECTION]: The documentation includes a "Prompt-Based Hooks" feature, which enables providing natural language instructions to the agent triggered by specific events. This creates a surface where automated inputs could influence subsequent agent behavior.
[SAFE]: Static detection identified destructive command patterns (e.g., 'rm -rf /', 'chmod 777') and a fork bomb pattern (':(){:|:&};:') in examples/hook-scripts.md and references/hooks-advanced.md. Analysis confirms these patterns are contained within a defensive validation script (validate-command.sh) and documentation examples designed specifically to detect and block such dangerous commands.
[SAFE]: External dependencies and MCP servers referenced in the documentation target well-known services (Stripe, GitHub) or official repositories maintained by the Model Context Protocol organization and the skill's author (JosiahSiegel).

advanced-features-2025