What problem does it solve? Prompt-level changes to squad coordinator and agent charter templates cannot be caught by unit tests because the prompts are interpreted by an LLM at runtime, so regressions only surface during real sessions. This Skill provides a repeatable end-to-end workflow for validating template changes against a locally built CLI before they ship. ## Core Features & Use Cases - Local build and link verification: Builds the squad SDK and CLI from your feature branch, links the CLI, and confirms the -preview version suffix so you test your changes rather than the published package. - Disposable test repo scenarios: Creates fresh git repos, runs squad init, executes real Copilot CLI sessions with --allow-all-tools, and verifies state landed in the expected location. - PR progress tracking and progressive verdicts: Posts a live tracking comment on the PR, updates it at every step boundary, and PATCHes each scenario verdict immediately so results survive connection drops. - Use Case: After editing squad.agent.md to add state-backend-aware spawn templates, run this Skill to build the CLI, spin up a test repo, run a coordinator session, and post a structured PASS/FAIL verdict table to the PR. ## Quick Start Ask the agent to run end-to-end validation of your squad template changes by building the local CLI, creating a disposable test repo, running a squad session, and posting the verdict to the PR.
