The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests conversation history and local files to update project documentation, creating an indirect prompt injection surface where external or user-provided data could be persisted to influence future sessions. 1. Ingestion points: Conversation history and local files such as CLAUDE.md or TODO. 2. Boundary markers: No explicit delimiters or ignore markers are mandated for the stored output. 3. Capability inventory: File-write access to the local project directory and optional, user-gated external tracker access. 4. Sanitization: No explicit content sanitization or escaping is described beyond the agent's logic for relevance.

stow