game-research
Warn
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructions direct the agent to write its own implementation for multiple Python scripts such as roblox_game.py, wiki_crawl.py, yt_search.py, yt_grab.py, strip.py, and img_study.py and subsequently execute them. This pattern of runtime code generation and execution poses a risk of arbitrary code execution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-generated content from YouTube transcripts and Fandom wikis. This content could contain malicious instructions designed to hijack the agent's context or analysis phase.
- Ingestion points: Untrusted data enters the agent context through the outputs of yt_grab.py (transcripts) and wiki_crawl.py (wiki articles) located in the references/ folder.
- Boundary markers: Absent. The skill does not define clear boundaries or 'ignore' prompts for the ingested text.
- Capability inventory: The agent utilizes subprocess calls (running the generated Python scripts) and file system write operations.
- Sanitization: Absent. There is no requirement or logic provided for sanitizing or validating the ingested text before the agent interprets it.
- [EXTERNAL_DOWNLOADS]: The skill workflow involves downloading data and files from various external sources. While some (Roblox API, YouTube, Fandom) are well-known, the Style study phase explicitly directs the agent to download 'originals' from arbitrary third-party source pages discovered via Google Images.
- [COMMAND_EXECUTION]: The skill utilizes shell commands to execute the generated Python scripts with the 'python -I' command and relies on external binaries like yt-dlp and ffmpeg for media processing.
Audit Metadata