Video Captioning
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses established media tools for transcription and follows a structured pipeline for reviewing and saving caption artifacts without suspicious behaviors or external calls to unknown domains.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from video files and generated transcripts.
- Ingestion points: Video content and transcript data retrieved via
media.caption_videoor host-local paths (SKILL.md). - Boundary markers: Absent. There are no instructions directing the agent to ignore or isolate potentially malicious content within the processed transcripts.
- Capability inventory: Tools include
media.caption_video,media.upload_file,media.check_result, andmedia.account_balance, along withworkspace-writepermissions for saving reports and receipt artifacts (SKILL.md). - Sanitization: Not present. The skill focuses on terminology and punctuation normalization but does not perform security-related sanitization or escaping of the input data.
Audit Metadata