design-system
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill utilizes a fetching tool to analyze content from external URLs provided by the user for visual style extraction, which serves as a potential vector for indirect instructions from untrusted sources to influence the agent's behavior.
- Ingestion points: Website URL analysis in Phase 0 (Reference Input) of
SKILL.mduses theWebFetchtool. - Boundary markers: Absent; the skill instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the fetched website content.
- Capability inventory: The skill has the capability to write files to the local and home directories (
SKILL.mdPhases 3 and 5) and invoke external image generation via thecreator-stack:nanobananaskill. - Sanitization: Absent; the skill performs visual property extraction on the raw content fetched from external domains without specific sanitization of textual instructions.
- [PERSISTENCE]: The skill provides an option to save design system documents and generated image assets to a hidden directory in the user's home folder (
~/.claude/.context/design-systems/). This feature is designed to maintain the design system's availability and context across multiple projects and future agent sessions.
Audit Metadata