skills/andabove/skills/prototype/Gen Agent Trust Hub

prototype

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting user prompts and project code to generate functional prototypes. 1. Ingestion points: User prompt and existing source code (SKILL.md). 2. Boundary markers: Absent in prompt construction. 3. Capability inventory: Generation of HTML, JS, and React components for local execution. 4. Sanitization: Not specified. The risk is mitigated by the instruction to use isolated modules and shareable single-file formats.
  • [DYNAMIC_EXECUTION]: The skill involves creating executable code based on user requirements. The instructions provide security guardrails by recommending in-memory state and stubs instead of real database connections or persistent storage.
  • [COMMAND_EXECUTION]: The skill directs the agent to ensure prototypes can be run using project task runners or interpreters like pnpm, bun, and python, which are standard development practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 11:30 AM
Security Audit — agent-trust-hub — prototype