define-architecture
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The validation loop defined in SKILL.md instructs the agent to execute shell commands such as
npm run lint,npm run check-types, andnpm run test, which involves subprocess creation as part of the project verification workflow. - [PROMPT_INJECTION]: The skill processes untrusted user data regarding project scope and codebase structure in SKILL.md to generate architecture briefs. It lacks explicit boundary markers or sanitization routines to isolate user-supplied context from the agent's instructions, creating a surface for indirect prompt injection that could potentially influence the generated recommendations or the subsequent validation commands.
Audit Metadata