vibe-build

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by relying on external data to drive its behavior.
  • Ingestion points: The agent reads the master plan from AGENTS.md and detailed specifications from agent_docs/ and docs/ PRD/TechDesign files.
  • Boundary markers: The skill lacks explicit delimiters or instructions to distinguish between trusted system instructions and potentially untrusted project documentation.
  • Capability inventory: The skill has broad capabilities including file system modification (implementing features) and command-line execution (npm test, npm run lint, npm install).
  • Sanitization: No validation or sanitization is performed on the content read from external markdown files before it influences agent actions.
  • [COMMAND_EXECUTION]: The workflow requires the agent to run various shell commands to verify the build and resolve errors.
  • Specific commands mentioned include npm test and npm run lint.
  • The skill suggests the agent can install new dependencies (e.g., npm install @supabase/auth-helpers-nextjs) if a build error is encountered. While these are standard developer tools, they are triggered by the contents of the AGENTS.md file, which is an untrusted ingestion point.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 12:20 PM