The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform automated software installation if dependencies are missing.
Evidence: Step 8 in SKILL.md lists Python 3.11, yt-dlp, ffmpeg, OpenCV, and ImageMagick as requirements and states: "If something is missing, attempt to install it yourself."
[DYNAMIC_EXECUTION]: The global guidelines allow for the creation and execution of temporary scripts to process API tasks.
Evidence: HOUSEKEEPING.md under 'Additional details' states: "You CAN write a throwaway script, e.g. to repeat one request over many items."
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its brand research phase, creating an injection surface.
Ingestion points: Researching the "look of the show" via artwork, websites, and existing clips (Step 9, SKILL.md).
Boundary markers: None specified for this step.
Capability inventory: File writes (preferences), tool execution (ffmpeg, yt-dlp, whisper.cpp), and API calls.
Sanitization: Not explicitly mentioned for external research data.
[PERSISTENCE]: The skill configures recurring tasks to maintain a daily workflow.
Evidence: Step 10 in SKILL.md and the 'Automation' section in HOUSEKEEPING.md describe setting up a "daily growth" loop using the fountain-daily-growth skill.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external sources, including the vendor's own repositories and specialized model files.
Evidence: Fetches plugins and configuration from the vendor's official GitHub repository (github.com/fountain-fm/fountain-skills).
Evidence: Downloads a whisper.cpp model file (ggml-base.en.bin) to ~/.cache/whisper (Step 8, SKILL.md).

fountain-onboarding