meta-muse-video-analysislisted

# Meta Muse Video Analysis Analyze one local video with `scripts/analyze_video.py`. Keep the model fixed; never add or substitute a model selector. ## Workflow 1. Resolve the user's video to an absolute local path. 2. Preserve the user's analysis prompt verbatim unless they ask for prompt improvement. 3. Confirm that `META_MUSE_KEY` exists without printing its value. 4. Run: ```powershell python "<skill-directory>\scripts\analyze_video.py" "C:\absolute\path\video.mp4" --prompt "Describe the requested analysis" ``` For long or multiline prompts, write the prompt to a temporary UTF-8 text file and use `--prompt-file`. Use `--output` when the user requests a saved result: ```powershell python "<skill-directory>\scripts\analyze_video.py" "C:\absolute\path\video.mp4" --prompt-file "C:\absolute\path\prompt.txt" --output "C:\absolute\path\analysis.md" ``` 5. Return the analysis or link the saved output. Surface API errors exactly enough to diagnose access, quota, format, or region issues, but never expose the API key. ## Operational rules - Read the API key only from `META_MUSE_KEY`. - Use only `muse-spark-1.2-contributor` through `https://api.meta.ai/v1`. - In Codex, run the script with network approval. The key may be visible only in the approved outside-sandbox process; never copy it into command arguments or output. - Check the model catalog before upload. If Contributor access is absent, stop without uploading and never fall back to standard `muse-spark-1.2`. - Upload