tasteroll
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes repository files (CSS, README, content) to inform its design audit and generation phases, creating a potential surface for instructions embedded in external data.
- Ingestion points: SKILL.md (Audit phase scans source CSS and repo files; Intake phase scans README and package.json).
- Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded prompts within the audited files.
- Capability inventory: The skill primarily generates markdown artifacts (DESIGN-SYSTEM.md). It does not demonstrate the capability to execute arbitrary shell commands or perform network operations within the provided scripts.
- Sanitization: Section 7 of SKILL.md ("Gate") defines security requirements for output, including an "Inert" requirement that explicitly forbids scripts in exported HTML, effectively mitigating the risk of persistent cross-site scripting (XSS) from processed data.
- [SAFE]: The local JavaScript engine (assets/tasteroll-engine.js) implements a standard, deterministic pseudo-random number generator (xoshiro128++) for reproducible design rolls. It contains no obfuscation or dangerous dynamic execution logic.
Audit Metadata