design-system-keeper
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill operates by scanning repository files such as CSS themes, Tailwind configurations, and component source code to extract design tokens and patterns. This ingestion of external, potentially untrusted repository data creates an indirect prompt injection surface.\n
- Ingestion points: The skill performs four sweeps across the repository as defined in 'references/extraction.md', reading tokens, components, and page layouts.\n
- Boundary markers: No specific delimiters or safety instructions are implemented to isolate the data read from the repository.\n
- Capability inventory: The skill is configured with 'mutating: true' and has write access to 'DESIGN.md', 'AGENTS.md', and the 'design/' directory.\n
- Sanitization: The skill lacks explicit sanitization or validation logic for the data extracted from the repository before using it to generate documentation.\n- [COMMAND_EXECUTION]: The skill execution includes simple shell commands to detect the presence of existing design documentation and git repository status. These commands are used for state detection and do not involve sensitive data or network access.
Audit Metadata