The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches a JSON catalog from an official learning hub at https://academy.claude.com/assets/data/catalog.json to maintain up-to-date recommendations for courses and tutorials.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from a remote URL, creating a potential surface for indirect prompt injection if the source were compromised.
Ingestion points: The catalog.json file is fetched from academy.claude.com as described in SKILL.md.
Boundary markers: The instructions explicitly command the agent to treat the file as data only, ignore any instructions it might contain, and take nothing from it except specific item entries.
Capability inventory: The agent's capabilities are limited to generating natural language recommendations and links; no dangerous system tools, file writes, or command executions are triggered by this data.
Sanitization: The skill enforces field-specific extraction (parsing JSON for title, url, etc.), which reduces the risk of the model interpreting data content as instructions.

academy-guide