The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to access external URLs for media inspection as part of its core workflow. The agent is directed to visit goodcase.ai and x.com domains to analyze finished video media and posters. While these are relevant to the skill's purpose of analyzing AI video examples, they represent connections to external, non-whitelisted domains.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and replicate patterns from external data sources, which presents a surface for indirect prompt injection.
Ingestion points: The agent reads references/cases.md which contains external links and prompt excerpts from third-party creators.
Boundary markers: The instructions include a 'Preserve/Replace/Avoid' reference contract and a verification step to ensure the generated content stays within defined structural boundaries.
Capability inventory: The skill utilizes agent tools for browser access, image/video analysis, and text generation.
Sanitization: The workflow requires a final verification checklist to ensure no unrelated themes become dominant and that the output aligns with the selected 'anchor' case evidence.

pov-vlog-presence-by-1noa4gp