The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several Python utility scripts such as scripts/check_graphify_install.py, sub-skills/graph-building/scripts/build_tiny_graph.py, and sub-skills/agent-integration/scripts/install_platform_probe.py. These scripts use the subprocess module to invoke the graphify CLI for smoke testing and installation verification. They implement defensive measures including the use of isolated temporary directories and the filtering of sensitive environment variables (e.g., API keys for Gemini, Anthropic, and OpenAI) from the execution environment.
[PERSISTENCE]: Integration workflows provided by the skill involve installing Git hooks (post-commit, post-checkout) and assistant-specific configuration files (e.g., for Claude Code, Cursor, and Gemini). These mechanisms serve to keep knowledge graphs synchronized with code changes and provide persistent guidance to AI agents. These actions are the primary purpose of the skill and are only executed via explicit user-facing commands.
[INDIRECT_PROMPT_INJECTION]: As the tool is designed to process untrusted external codebases and URLs into graph format, it possesses an inherent indirect prompt injection surface. The skill instructions mitigate this by directing the agent to citation-based answering from graph evidence and providing clear guidance on handling ambiguous or missing data without hallucinating relationships.
[DYNAMIC_EXECUTION]: The scripts/check_graphify_install.py script uses importlib.import_module to programmatically verify the presence of the graphify package and its submodules. This is a standard diagnostic practice and does not involve the execution of untrusted input.

graphify