fullstack-dev
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of full-stack applications by gathering user requirements (SKILL.md, Step 0), which introduces an attack surface where the resulting application processes untrusted input. However, the skill provides strong mitigations.
- Ingestion points: User-supplied architectural requirements and configuration inputs in Step 0.
- Boundary markers: Explicit instructions in SKILL.md and core principles define that the agent must "trust nothing from client" and use boundary delimiters.
- Capability inventory: The skill defines patterns for API endpoints, database interactions, and network operations across TypeScript, Python, and Go scripts.
- Sanitization: The skill mandates input validation on all endpoints using Zod, Pydantic, or Go validators.
- [SAFE]: The skill demonstrates proactive security measures, such as instructing the agent to use environment variables for secrets, implementing secure JWT authentication with httpOnly cookies, and configuring CORS with explicit origin lists rather than wildcards. No obfuscation, unauthorized command execution, or data exfiltration patterns were found.
Audit Metadata