The Agent Skills Directory

[PROMPT_INJECTION]: Potential risk of indirect prompt injection from scraped ad content. \n * Ingestion points: Facebook Ad Library (scraped via Apify actor curious_coder~facebook-ads-library-scraper or Browser MCP). \n * Boundary markers: A dedicated warning quote block is present at the beginning of the skill instructions to establish context isolation. \n * Capability inventory: The skill can trigger Apify actor runs, execute local media processing commands (ffmpeg, whisper.cpp), and write data to Airtable. \n * Sanitization: Includes a specific negative constraint instructing the agent to quote and stop if external content attempts to address or command the AI directly. \n- [EXTERNAL_DOWNLOADS]: Fetches data and media from the Meta Ad Library via the Apify platform and utilizes a remote analysis tool hosted at https://mcp.arcads.ai. These represent functional dependencies required for the skill's primary purpose. \n- [COMMAND_EXECUTION]: Relies on local installations of ffmpeg and whisper.cpp to process video files and generate transcripts.

spy-ads