The Agent Skills Directory

[COMMAND_EXECUTION]: The script uses subprocess.run() to execute ffmpeg for extracting audio from downloaded video files. This is a standard utility for media processing and the arguments are statically defined, minimizing risk.
[EXTERNAL_DOWNLOADS]: The skill fetches content from x.com and twimg.com (official X domains) and allows downloading video/image assets to the local system. These are well-known services and the downloads are essential to the skill's primary purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external X posts, which could contain malicious instructions.
Ingestion points: scripts/fetch_tweet.py reads content from X URLs or fallback JSON/text files.
Boundary markers: The skill uses YAML frontmatter and markdown headers to structure the output, but does not explicitly warn downstream agents to ignore instructions within the ingested text.
Capability inventory: File system writes (Markdown and images) and subprocess calls for media processing.
Sanitization: The script performs basic sanitization of filenames but renders the tweet body directly into Markdown.
[SAFE]: References to external AI models (SenseVoice) and packages like funasr and modelscope target legitimate open-source repositories and registries.

x-ingest