The Agent Skills Directory

[COMMAND_EXECUTION]: The skill installs a Git post-commit hook through the graphify hook install command (documented in references/hooks.md). This persistence mechanism automatically triggers the graph rebuild pipeline after every commit, modifying the local repository configuration.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs external Python dependencies, specifically the graphifyy package, using uv or pip during its setup phase (SKILL.md Step 1). It also facilitates cloning arbitrary GitHub repositories and fetching content from external URLs for ingestion into the knowledge graph.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to its document processing architecture:
Ingestion points: Files are ingested from a user-provided directory (INPUT_PATH) and from content cloned from GitHub repositories or downloaded from arbitrary URLs.
Boundary markers: The extraction prompt (found in references/extraction-spec.md) defines structural rules for graph extraction but lacks explicit delimiters or instructions to the subagent to ignore potentially malicious directives contained within the source data.
Capability inventory: Subagents are explicitly invoked with subagent_type="general-purpose", which grants them Bash and Write tool access to handle chunk processing (SKILL.md Step 3B).
Sanitization: No evidence of content sanitization or filtering is present before passing file data to the subagents for semantic analysis.
[COMMAND_EXECUTION]: The skill employs complex shell scripts to manage its execution environment, including dynamic discovery of the Python interpreter and the execution of Python-based extraction logic using the host's shell commands.

graphify