ideate
Pass
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill enforces a strict conversational protocol for architectural planning, requiring the agent to ask clarifying questions before proposing designs. This behavior prevents jumping to unvalidated conclusions.
- [COMMAND_EXECUTION]: The skill uses git commands (
git add,git commit) to version design specifications within thedocs/arc/specs/directory. These operations are appropriate for the skill's purpose and are limited to documentation management. - [PROMPT_INJECTION]: A surface for indirect prompt injection exists because the skill reads local files such as
docs/vision.mdanddocs/arc/progress.md. Boundary markers are maintained through the use of structured tool calls for user interaction, and the skill's capabilities are restricted to documentation creation, minimizing potential impact.
Audit Metadata