design-craft-library
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The EXTRACT mode processes content from external URLs provided by users. This ingestion point is susceptible to indirect prompt injection where a malicious website could attempt to influence the agent's behavior during the drafting phase.
- Ingestion points: SKILL.md (EXTRACT mode fetches external URL content).
- Boundary markers: SKILL.md (Step 5: 'STOP. Show every proposed entry in full and wait for approval').
- Capability inventory: SKILL.md (File writing to entries/ and references/index.md, network fetching via Playwright/Puppeteer or direct read).
- Sanitization: SKILL.md (Instruction to extract only portable, brand-detached moves and exclude sensitive data like client names or internal tokens).
- [EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch the rendered DOM and screenshots of external websites during the pattern extraction process. Evidence is located in SKILL.md under the EXTRACT section.
Audit Metadata