fountain-onboarding
Warn
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to install a plugin from the author's GitHub repository at
https://github.com/fountain-fm/fountain-skills. It also involves downloading machine learning model files for whisper.cpp to the local~/.cache/whisperdirectory. - [COMMAND_EXECUTION]: The instructions authorize the agent to automatically install missing system-level software, including Python 3.11, yt-dlp, ffmpeg (with specific modules like libass and fontconfig), OpenCV, and ImageMagick. This requires the execution of shell commands and package managers.
- [DYNAMIC_EXECUTION]: Guidelines in
HOUSEKEEPING.mdpermit the agent to write and execute "throwaway scripts" to handle repetitive API requests. While it mandates deleting these scripts after use, it establishes a pattern of generating and running code at runtime. - [PERSISTENCE]: The skill includes steps to set up "automatic daily growth," which implies the creation of recurring scheduled tasks or background processes to automate Fountain operations on the host machine.
- [INDIRECT_PROMPT_INJECTION]: The skill researches external websites, show artwork, and video clips to generate brand guidelines and clip styles, creating an attack surface for instructions embedded in external media.
- Ingestion points: External websites, show metadata, and existing video clips processed in Step 9 of
SKILL.md. - Boundary markers: No specific delimiters are used to separate researched content from agent instructions.
- Capability inventory: The skill can install software, write files and temporary scripts, and modify project settings via API.
- Sanitization: No explicit sanitization or validation of the data retrieved from external research is performed before it is used to influence project preferences.
Audit Metadata