The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch data from external URLs defined in the _meta.sources field of its data file and use that content to update local storage.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where it consumes and writes data from external, non-hardcoded URLs to the local filesystem. A lack of sanitization or validation instructions for this external content could allow a compromised source to inject malicious data or instructions into the agent's environment.
Ingestion points: External URLs referenced in _meta.sources and the data/points-valuations.json file.
Boundary markers: None provided in instructions for handling the external data.
Capability inventory: Performs network requests (fetch) and local file writes (update file) as defined in SKILL.md.
Sanitization: No validation or escaping logic is defined for the downloaded content.
[TIME_DELAYED_CONDITIONAL]: The skill includes logic that gates network operations and file system modifications behind a 45-day staleness check based on the _meta.last_updated metadata field.

points-valuations