artisan
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill exhibits proactive security awareness by explicitly instructing the agent to avoid React versions vulnerable to CVE-2025-55182 (React2Shell) and prohibiting the storage of sensitive data on the client side.
- [SAFE]: All external documentation links and library references point to well-known technology organizations and official framework repositories, including those for React, Next.js, Vercel, Svelte, Vue, and the W3C.
- [INDIRECT_PROMPT_INJECTION]: The skill processes prototype data from upstream agents (e.g., Forge), creating a surface for indirect prompt injection. This is assessed as a low risk inherent to code-generation tasks, with no evidence of malicious exploitation.
- Ingestion points: The ANALYZE phase in SKILL.md reads Forge prototypes and user-provided requirements.
- Boundary markers: The workflow defines a clear separation between analysis, design, and implementation phases, although specific data delimiters are not mandated.
- Capability inventory: The skill is restricted to implementing frontend components and state management; it does not include capabilities for arbitrary shell execution or network exfiltration.
- Sanitization: No explicit input sanitization is mentioned, which is consistent with its primary function as a development aid.
Audit Metadata