enhance-prompt

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Prompt Injection] (SAFE): No instructions attempting to bypass safety filters or override agent behavior were detected. The instructions focus entirely on prompt enhancement logic.
  • [Data Exposure & Exfiltration] (SAFE): The skill utilizes the 'Read' tool to access 'DESIGN.md' within the project context. It does not attempt to access sensitive system paths (e.g., ~/.ssh, .env) or perform network operations.
  • [Remote Code Execution] (SAFE): The skill does not define external dependencies or execute remote scripts. The installation command provided in the README refers to a trusted source (google-labs-code).
  • [Indirect Prompt Injection] (LOW): The skill features an ingestion surface for untrusted data from the 'DESIGN.md' file.
  • Ingestion points: SKILL.md reads 'DESIGN.md' content to extract design system blocks.
  • Boundary markers: The content is wrapped in a 'DESIGN SYSTEM (REQUIRED)' section in the output, providing some context to the downstream agent.
  • Capability inventory: The skill has 'Read' and 'Write' permissions, which are used to generate text prompts or write them to local files.
  • Sanitization: No explicit sanitization or escaping of the 'DESIGN.md' content is performed before interpolation into the prompt.
  • Note: Because this is the primary intended use-case of the skill, the severity is categorized as SAFE.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:42 PM