app-builder
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to take natural language requirements from a user and translate them into code and project structures. This ingestion of external data represents a surface for indirect prompt injection attacks.
- Ingestion points: User prompts are analyzed by the orchestrator (SKILL.md) and the keyword matrix (project-detection.md).
- Boundary markers: No explicit delimiters are used to separate user-provided descriptions from the agent's internal instructions.
- Capability inventory: Specialist agents coordinated by this skill have the ability to create and modify files, install packages, and configure development environments.
- Sanitization: The skill does not implement specific filtering or sanitization of user input, relying instead on the model's inherent safety guardrails.
- [EXTERNAL_DOWNLOADS]: The project templates recommend the installation of numerous third-party packages from official repositories like NPM and PyPI. These include industry-standard tools such as Prisma, FastAPI, NextAuth, and Stripe SDKs. These downloads are associated with well-known and trusted services and registries, posing no unusual risk.
Audit Metadata