The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to detect project environments and manage Python dependencies via uv or pip. This is standard for developer-oriented tools and includes logic to safely identify the correct Python interpreter.
[EXTERNAL_DOWNLOADS]: The skill can optionally download GitHub repositories for analysis using graphify clone. These downloads are user-initiated via explicit CLI arguments.
[DATA_EXPOSURE]: The skill performs local file system reads to extract metadata and relationships from a codebase. It includes explicit 'Honesty Rules' and an 'audit trail' mechanism to differentiate between extracted facts and inferred relationships.
[REMOTE_CODE_EXECUTION]: Semantic extraction is performed by dispatching 'general-purpose' subagents. These subagents operate on local file chunks to produce JSON graph fragments. There is no evidence of arbitrary remote code execution outside of the intended graph extraction logic.
[CREDENTIALS_UNSAFE]: The skill mentions reading GEMINI_API_KEY or GOOGLE_API_KEY from the environment for semantic extraction. It correctly avoids hardcoding these keys and instructs users on standard environment variable usage.

graphify