prototype

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides instructions for creating 'throwaway' prototypes. It explicitly warns against using these for production and encourages isolation of logic from the host application.
  • [COMMAND_EXECUTION]: The instructions mention using existing project task runners (e.g., pnpm, python, bun) to run the prototypes. This is standard developer workflow and does not introduce new or unexpected command execution capabilities.
  • [DATA_EXPOSURE]: The skill recommends using in-memory state or local scratch files for prototypes to avoid affecting production databases. It does not contain patterns for exfiltrating sensitive data like credentials or SSH keys.
  • [PROMPT_INJECTION]: No evidence of prompt injection, jailbreak attempts, or instructions to ignore safety filters was found in the skill or its references.
  • [EXTERNAL_DOWNLOADS]: The skill does not instruct the agent to download or execute code from external or untrusted URLs. It relies entirely on the host project's existing environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 06:03 PM
Security Audit — agent-trust-hub — prototype