video-overlay

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to clone and use an external repository (https://github.com/agamm/video-agent), which provides the underlying toolkit for the video processing operations.
  • [COMMAND_EXECUTION]: The instructions involve running complex shell commands using uv run, ffmpeg, and Python scripts. These commands manipulate files and execute logic based on user input and external data.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from video transcripts (via transcribe --words) and uses it to determine the timing and content of overlays. 1. Ingestion points: Video files and transcript output referenced in SKILL.md. 2. Boundary markers: No delimiters or warnings are used when processing transcript data. 3. Capability inventory: Subprocess calls to ffmpeg and uv run, as well as file writes via PIL. 4. Sanitization: No content validation or sanitization is mentioned for the transcript-derived inputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 12:32 PM