app-builder
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests natural language prompts to generate application code and structures. This inherent attack surface is managed through a mandatory planning phase and a structured orchestration pipeline that includes a security auditing step.
- Ingestion points: Natural language requests for project creation as described in SKILL.md and project-detection.md.
- Boundary markers: Defined tech-stack mappings and directory scaffolding rules in scaffolding.md and TEMPLATE.md files.
- Capability inventory: The skill utilizes Write, Edit, and Bash tools for project setup.
- Sanitization: Employs a multi-agent verification process requiring a PLAN.md file as described in agent-coordination.md.
- [EXTERNAL_DOWNLOADS]: Project setup steps in the various templates include standard commands to fetch well-known frameworks and libraries from official package registries. These are necessary components for the stated development purpose and originate from trusted ecosystems (NPM, PyPI, Dart Pub).
- Evidence: Templates for Next.js, Astro, FastAPI, and others instruct the agent to use standard package managers to install common development dependencies.
Audit Metadata