The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a find command to discover the path of a script named graph_query.py in the ~/.claude/plugins directory and then executes it using python3. This runtime determination of an executable's path is a dynamic execution pattern.
[PROMPT_INJECTION]: The skill's fallback logic interpolates user-provided function names into shell commands (rg or grep) without sanitization. This creates an indirect prompt injection surface for command injection. 1. Ingestion points: Function names or entry points from user input in SKILL.md. 2. Boundary markers: None present. 3. Capability inventory: Execution of rg and grep shell commands in SKILL.md. 4. Sanitization: No input validation or escaping of the user-provided function name before shell interpolation.

call-chain