plan-to-goal
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input (rough plans) to generate autonomous goal specifications. It implements a robust mitigation strategy through a two-phase workflow separated by a 'Gate.' This ensures that all interpreted completion conditions, file-system boundaries, and execution constraints are explicitly reviewed and confirmed by the human user before a final goal prompt is emitted.
- [EXTERNAL_DOWNLOADS]: Documentation within the skill guides users to install the tool using a command that fetches content from the author's official GitHub repository (
github.com/leoluyi/skills). This is recognized as a standard vendor-controlled resource for skill distribution. - [COMMAND_EXECUTION]: The skill instructions prompt the agent to define completion criteria using shell commands (e.g.,
npm test,grep,typecheck). These are treated as descriptive examples for the 'Verification' section of a goal spec and are not executed by the agent during the skill's reasoning phase.
Audit Metadata