Use this process before claiming a TUI, chat, or agent-mode daily-driver issue is fixed.
- Identify the user-visible claim.
- Identify the active command path.
- Add or run an automated test for that path.
- Run a manual smoke step if terminal behavior is involved.
- Update known issues and current status.
- Link the proof from the ticket plan or fix log.
| Surface | Automated proof | Manual proof |
|---|---|---|
teaagent chat |
Parser/handler/controller tests. | REPL visible answer, cost, undo. |
teaagent tui |
Headless TUI command tests. | Terminal prompt, panel, command output. |
teaagent agent run |
CLI handler/run-store/audit tests. | Run id, summary, changed files. |
| Approvals | Approval manager and CLI/TUI prompt tests. | Exact scope is understandable. |
| Resume/review | Run-store rehydration tests. | Human can continue or review from run id. |
For commands advertised as dry-run or read-only:
- Create a fresh temp workspace.
- Snapshot file tree.
- Run the command.
- Snapshot file tree again.
- Fail if
.teaagentor other state appears without explicit initialization wording.
Use these terms consistently:
Claimed: agent or docs say it happened.Observed: logs or files show it happened.Verified: test or manual smoke proved it.Not tested: no direct proof yet.
Do not mark a daily-driver item fixed if any of these are true:
- The test bypasses the public path.
- The docs still warn about the supposedly fixed behavior.
- The terminal output uses stale lifecycle wording.
- The run evidence cannot be found by run id.
- The fix changes approval, root, cost, or undo behavior without human review.