The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from third-party videos and transcripts to generate resource content. This is identified as an attack surface, but risk is minimized through a mandatory human-in-the-loop approval step (Step 4) before the resource is built or published.
[EXTERNAL_DOWNLOADS]: Fetches video data from well-known services including YouTube, Instagram, and Apify. These downloads are performed to retrieve source material for the transcription process.
[COMMAND_EXECUTION]: Employs standard media and development tools such as ffmpeg, ffprobe, git, and gh to process video files and manage repository updates.
[DATA_EXPOSURE]: Uses configuration placeholders for sensitive identifiers like Supabase project IDs and Airtable base IDs, avoiding the use of hardcoded credentials.
[DATABASE_INTERACTION]: Executes SQL queries via a Supabase MCP tool. It uses dollar-quoting ($vault$) to wrap dynamically generated markdown content, which helps prevent SQL syntax errors or injection during the upsert process.

lead-magnet