figure-it-out
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading project configurations and monitoring pull requests, creating a surface where malicious instructions embedded in external data could influence the agent.
- Ingestion points: The skill monitors pull request status through
watch_pull_requestand reads project roles fromkickoff.yaml. - Boundary markers: The instructions mandate decomposing work into "atomic, independently-landable units" and using isolated briefs for task delegation (
delegate_task), which serves as a structural control boundary. - Capability inventory: The skill utilizes task orchestration and management tools including
delegate_task,t3_thread_launch,schedule_task, andtask_status. - Sanitization: The skill employs a verification discipline in "Phase C" that requires "inspecting the artifact, never a self-report," ensuring that results are validated against actual file changes rather than potentially manipulated status reports.
Audit Metadata