skills/leonvanzyl/skills/start-an-app/Gen Agent Trust Hub

start-an-app

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches official documentation and best-practice guidelines (agent skills) from trusted GitHub organizations including Vercel, Anthropic, and Shadcn. These are used to ensure the generated application adheres to the latest security and implementation standards of the underlying libraries.
  • [COMMAND_EXECUTION]: To fulfill its primary purpose, the skill executes standard development tools such as create-next-app, pnpm, and docker-compose. These commands are used to initialize the project, manage dependencies, and set up local development databases.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user requirements through an interview process. While this creates an attack surface for indirect prompt injection, the skill mitigates this risk by requiring the agent to generate a "build sheet" for explicit user approval before any code is written or executed. This creates a clear human-in-the-loop security boundary.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes a system visibility and operations dashboard to help the user monitor the app's behavior (e.g., background jobs, sent emails). This dashboard is designed with strict security controls, such as requiring administrator privileges and explicitly forbidding the rendering of secrets or API keys in the interface.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 07:07 PM
Security Audit — agent-trust-hub — start-an-app