The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or persistence mechanisms were detected. The skill is designed for automated video production and follows legitimate operational patterns.
[COMMAND_EXECUTION]: The skill uses subprocess.run() extensively across multiple scripts (e.g., scripts/render_full.py, scripts/build_assets.py) to invoke ffmpeg and ffprobe. This is the intended primary purpose of the skill for video rendering and metadata extraction.
[EXTERNAL_DOWNLOADS]: The skill references legitimate external dependencies and tools, including whisper.cpp for local transcription and the video-use helper repository from GitHub. It also utilizes the Higgsfield CLI (installed via NPM) for image generation. All downloads target well-known or author-specified official sources.
[CREDENTIALS_SAFE]: The skill demonstrates safe credential management. Instructions in SKILL.md and references/PUBLISHING.md explicitly forbid hardcoding API keys, directing users to store them in environment variables or secure key files (e.g., ZERNIO_API_KEY, ELEVENLABS_API_KEY).
[DATA_EXPOSURE]: Network operations are restricted to the primary service providers required for the pipeline: zernio.com for publishing and higgsfield.ai for thumbnail generation. No evidence of unauthorized data harvesting or exfiltration was found.
[INDIRECT_PROMPT_INJECTION]: The skill processes transcription data (ASR) to generate captions and thumbnail prompts. While this represents a data ingestion surface, the risk is mitigated as the content is used for visual generation rather than executable logic. The scripts/corrections.py file provides a sanitization layer for ASR domain terms.

longform-to-content