figmalint-design-system-auditing
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests Figma component metadata (layer names, descriptions, and properties) which are external and potentially untrusted. This content is interpolated into prompts for LLM providers (Claude, OpenAI, Gemini), presenting a surface for indirect prompt injection.
- Ingestion points: Component metadata is ingested via
analyzeComponentNodeandanalyzeTokensinSKILL.md. - Boundary markers: The provided documentation does not show explicit delimiters or "ignore instructions" warnings when interpolating component data into AI prompts.
- Capability inventory: The skill calls AI providers and can perform network requests to sync data to external documentation platforms via
fetch. - Sanitization: There is no evidence of sanitization or content filtering for component data before it is sent to the LLM.
- [EXTERNAL_DOWNLOADS]: The documentation references external sources for plugin installation and development setup.
- Evidence: Installation instructions include a link to Figma Community and a command to clone a repository from
https://github.com/southleft/figmalint.git.
Audit Metadata