podcast-edit
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes transcription data generated from user-provided audio or video files, creating an attack surface where instructions embedded in the speech could influence the agent during the scanning phase.\n
- Ingestion points: The agent is instructed in SKILL.md to scan JSON transcriptions for start markers, end markers, and other content indicators.\n
- Boundary markers: No explicit delimiters or warnings are used to instruct the model to ignore instructions embedded within the transcribed text.\n
- Capability inventory: The skill possesses significant capabilities including file system access, complex shell command execution via ffmpeg, and network access via curl to external APIs.\n
- Sanitization: There is no evidence of sanitization or filtering of the transcription text before the agent processes it for decision-making.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes well-known external services and packages to perform its tasks.\n
- Evidence: Fetches audio transcriptions from the official OpenAI Whisper API (api.openai.com).\n
- Evidence: Mentions the resemblyzer Python package for speaker diarization.
Audit Metadata