The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from multiple sources, creating a vulnerability surface for indirect prompt injection.
Ingestion points: The skill fetches and processes content from YouTube URLs, audio files, PDFs, book scans, screenshots, and GitHub repositories (Phase 1, SKILL.md).
Boundary markers: The skill lacks explicit instructions or delimiters to warn the model to ignore potential instructions embedded within the processed media.
Capability inventory: The skill uses put_page, add_link, add_timeline_entry, and file_upload to persistently modify the agent's "brain" (knowledge base) and filesystem.
Sanitization: There are no explicit sanitization routines to filter out or escape natural language instructions that might be hidden in OCR text, transcripts, or repository files (like READMEs).

media-ingest