The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process live documentation from external sources (docs.medusajs.com) using WebSearch and WebFetch tools.
Ingestion points: Data is ingested via external web searches and fetches defined in SKILL.md to ensure up-to-date security information.
Boundary markers: None present; the skill treats documentation content as instructions for the agent.
Capability inventory: The skill uses Write, Edit, and Bash tools which could be influenced by external input.
Sanitization: No explicit sanitization of fetched documentation is implemented, relying on the integrity of the documentation source.
[EXTERNAL_DOWNLOADS]: The skill fetches updated implementation details and middleware signatures from the official Medusa documentation site (docs.medusajs.com). This is a well-known service associated with the skill's core purpose.
[DATA_EXPOSURE]: The skill advocates for secure secret management, specifically instructing the user to use environment variables (COOKIE_SECRET, JWT_SECRET, DATABASE_URL) and cryptographically random strings, rather than hardcoding credentials.

medusa-security