prompt-builder
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFE
Full Analysis
- [Sandbox-Constrained Operations]: The skill implements a deterministic sandbox contract for all execution and testing activities. By confining file edits and test runs to specific subfolders within
.copilot-tracking/sandbox/, it ensures that prompt development work is isolated from the primary codebase.\n- [Subagent Orchestration]: The workflow relies on specialized subagents (Tester, Evaluator, Researcher, Updater) to perform scoped tasks. This modular approach follows a defined orchestration matrix that governs inputs and outputs, maintaining clear boundaries between different stages of the prompt-building process.\n- [Automated Cleanup]: The skill includes explicit instructions to remove temporary sandbox artifacts and folders created during the execution loop before providing a final response. This practice supports workspace hygiene and prevents the persistence of transient test data.
Audit Metadata