arojinle
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is entirely instructional, defining a workflow for the agent to resolve complex choices through user interviews. It does not contain any shell commands, file system operations, or network requests. References to external tools such as 'iwadi', 'oro', and 'html-artifact' are consistent with standard project-scoped capabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input as part of its core interview functionality, which serves as an ingestion point. However, the risk is minimal as the skill implements a structured question-and-answer format and mandates a final confirmation step to ensure shared understanding before concluding.
- Ingestion points: User responses to the decision-tree interview questions (SKILL.md).
- Boundary markers: Employs a specific question card template to structure user interactions.
- Capability inventory: References research ('iwadi'), drafting ('oro'), and visualization ('html-artifact') tools, which are limited to project-specific contexts.
- Sanitization: Relies on a multi-round 'frontier' validation logic that forces re-evaluation of the decision tree based on confirmed answers.
Audit Metadata