The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses multiple shell commands to handle media processing, including yt-dlp for downloads, ffmpeg for frame extraction, ffprobe for metadata, and curl for API requests.
[EXTERNAL_DOWNLOADS]: The skill downloads content from Instagram, TikTok, and YouTube, and relies on Apify actors for scraping social media data.
[CREDENTIALS_UNSAFE]: The skill refers to an OPENAI_API_KEY expected to be present in a .env file, which is a standard practice for managing secrets but marks the skill's dependence on external credentials.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from video transcripts, captions, and on-screen text.
Ingestion points: Data is collected from external URLs via scraping tools and transcription services in references/shortform.md, references/carousel.md, and references/longform.md.
Boundary markers: A clear warning is present in SKILL.md instructing the agent to treat all scraped content as data, not instructions.
Capability inventory: The skill uses subprocess execution (shell commands) and network communication.
Sanitization: The instructions explicitly command the agent to stop and quote any scraped text that appears to be an instruction or address the agent directly.

analyze-content