video-overlay-cleanup-agent

• Verdict: Pass
• Confidence: High
• Analyzed: Apr 12, 2026, 9:44 AM
• Findings: 0
• Version: 1.0.1

Overview

The skill's code, declared requirements, and runtime instructions are consistent with a video-overlay cleanup workflow that optionally calls a Gemini image-editing API; nothing requested or installed appears unrelated to that purpose.

This skill appears internally coherent for cleaning video overlays, but consider the following before installing or running: 1) GEMINI_API_KEY is required for per-frame generative edits — use an account with appropriate billing limits because per-frame API calls can be expensive. 2) The frame-edit path executes an 'editor command' for every frame; ensure the editor_cmd you supply (or the provided gemini wrapper) is trusted because restore_frames.py runs that command via subprocess. 3) The skill manipulates local files and will create and (by default) delete a workdir; run it in a controlled directory and use --keep-workdir while testing. 4) Ensure google-genai and Pillow are installed in your environment before using Gemini mode. 5) Inspect the scripts (they are included) and test with short/clipped videos first to validate quality and cost. If you need stricter isolation, run the skill in a sandboxed environment or with a limited API key and low concurrency.