test-iterationlisted

# Feature Test Iteration **Branch:** `$ARGUMENTS` (if not provided, stop and ask for the branch). This skill orchestrates a full QA cycle and produces a verdict: ✅ GO / ⚠️ GO with deferrals / ⛔ NO-GO. It delegates review to the `branch-review` skill and research to the `qa-research` skill, and uses bundled scripts for the deterministic git/deploy steps. The outcome is a tested feature plus a report. ## Principles (important) - **Do everything yourself that you technically can.** git, review, research, driving the UI by hand, calling the API as a client, reading logs/status, and using the project's own tooling on the non-prod env — `tinker`, `artisan` commands, DB changes, queries — to set up state, run the feature, and observe results. All of that is fine. Turn to the user at step 7 (approval). - **The one hard boundary: never put a file onto the server.** No `scp`, no `docker cp`, no writing/deploying a script, harness, or probe **file** to the server or container — not even read-only, not even to set up a clean state. **Everything else a tester does is allowed**: run `tinker`/`artisan`, change DB state, query the DB, run commands **directly** on the non-prod env. If you need to run some logic, run it as a direct command (e.g. `artisan tinker --execute=…`), **don't upload a file to do it.** (Production is separately off-limits — see the production boundary below.) - **Prefer deterministic tools over guessing.** Use the bundled scripts for git diff and deploy verification