ingest
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external URLs and video transcripts to update a persistent vault. Ingestion points: Content from external URLs and transcripts fetched via yt-dlp. Boundary markers: No explicit protective delimiters or instructions to ignore embedded commands are used during the summary or integration steps. Capability inventory: The skill has file system access to read/write to the vault and can execute shell scripts. Sanitization: Basic HTML and text cleaning is performed by the script, but no instruction-level sanitization is present.
- [COMMAND_EXECUTION]: The skill invokes a local bash script 'scripts/fetch_transcript.sh' which executes 'yt-dlp' and 'jq' to extract video metadata and captions.
- [EXTERNAL_DOWNLOADS]: The skill fetches data from remote web services and video platforms to generate transcripts and summaries.
Audit Metadata