forge
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill contains comprehensive security guardrails, specifically addressing the risks of AI-assisted development. It mandates a 'COOL' phase for security spot-checks and prohibits delegating authentication, encryption, or payment processing to AI scaffolding tools.\n- [PROMPT_INJECTION]: The skill integrates code from external AI services, creating an indirect prompt injection surface. This is addressed with the following security measures:\n
- Ingestion points: Processes output from v0, Bolt.new, and Cursor (SKILL.md, references/ai-assisted-prototyping.md).\n
- Boundary markers: Uses workflow stages like SCAFFOLD and COOL to isolate and verify integrated code.\n
- Capability inventory: Capable of creating files and running local prototype servers (references/backend-poc.md).\n
- Sanitization: Instructions require the agent to scan for SQL/command injection, hardcoded secrets, and client-side auth bypasses before handoff (references/ai-assisted-prototyping.md).\n- [COMMAND_EXECUTION]: The skill provides templates for running local development servers (Express, Fastify) and tools like json-server for API mocking. These are documented as prototype-only tools and do not involve the execution of arbitrary or unauthorized commands.
Audit Metadata