troubleshooting-applisted

## Overview See `docs/roles.md` for the cast of characters and the authority framework this skill operates under — not redefined here. No worktree opened yet, no fix made yet — the cause is unknown, so there's nothing to branch for and nothing to patch. Diagnosis (Task 3) is read-only either way; whether it stays solo in this session or dispatches to a worker rooted in the app is `boss-say`'s execution-tier judgment. Once root cause is known, hand off to `boss-say` for the actual fix, which triages it and makes its own tier call. ## Task 1: Resolve the app Invoke `work-on` now. Even for a live incident, you need the target app before digging in. **Verification:** the target app is established before Task 2. ## Task 2: App-code or infrastructure? Before debugging application code, rule out an infrastructure cause: deployment failure, orchestrator/job state, DB connectivity, config/secret drift. Symptoms like "works locally, fails in an environment", "worked yesterday, no code changed", or "500 with no recent deploy to this app" point at infrastructure, not code. - **Looks infra-level:** hand off to this project's infrastructure/ops skill or team if one exists; otherwise tell the user plainly that this looks infrastructure-level and stop — don't debug application code for a problem that isn't in the code. - **Looks app-level, or infra is ruled out:** continue to Task 3. **Verification:** you can state which side the evidence points to, and why, before choosing a diagno