video-pipeline

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill initiates multiple processing stages using shell commands. It executes Python and Bash scripts including transcribe.py and batch-clip-v4.sh using nohup to manage long-running background tasks for video transcription and content clipping.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. It ingests untrusted data in the form of transcripts generated from external video files which are then used as prompts for downstream skills like insight-extractor and article-forge. This could allow instructions hidden in the audio to influence the agent's summary or article generation.
  • Ingestion points: Reads transcript.txt and transcript.json derived from external video files (SKILL.md).
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when passing transcript data to subsequent stages.
  • Capability inventory: The pipeline executes subprocesses via python3 and bash and performs file writes to the local workspace.
  • Sanitization: No sanitization or validation logic is defined for the transcribed text before it is processed by the LLM stages.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 09:02 AM