The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from web searches and fetched pages. Ingestion points: The research process in references/research.md fetches content from external URLs. Boundary markers: references/research.md explicitly instructs the agent to treat page text as data and avoid following embedded directions. Capability inventory: The skill can read and write files (brief.md, research.md, sources.json, topics.json) and perform web searches. Sanitization: Employs prompt-level instructions to mitigate risks from external content.
[EXTERNAL_DOWNLOADS]: The skill performs web searches and page fetches to research video topics, which is an expected behavior for its core functionality.
[COMMAND_EXECUTION]: The skill references a local Node.js script (scripts/check.mjs) for validating output files. The script is limited to basic file system reads and format validation.

24fps-brief