longform-clip-extract
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses an entrypoint to execute a TypeScript-based harness (
scripts/harness/longform-clip-extract.ts) via Node.js to process media files. This is standard behavior for automated video editing tasks and is performed within the local environment.\n- [PROMPT_INJECTION]: The skill ingests data from external JSON files (approval and timestamp data), which represents a surface for indirect prompt injection. This risk is localized to the media extraction parameters and is mitigated by human-in-the-loop approval workflows mentioned in the documentation.\n - Ingestion points: SKILL.md inputs (approvalPath, boundarySnapPath, timestampsPath)\n
- Boundary markers: No specific delimiters or "ignore" instructions are present for the ingested metadata files.\n
- Capability inventory: The skill utilizes the shell, read, and write tools as specified in its YAML frontmatter to manage media files and process execution.\n
- Sanitization: A validation checklist is provided to ensure the agent verifies that extracted clips are non-empty and that timestamps are within the correct ranges.
Audit Metadata