End-to-end example
Follow the recorded Taskboard exercise from missing behavior to reviewed evidence.
This is a condensed account of the local Codex exercise in examples/taskboard-session/, originally performed as DevMap. The map now lives in .devthread/; its original reports are preserved. The commands below use the renamed scripts. Conversation prompts shown on the landing page are illustrative summaries.
Starting point
Taskboard is a small JSON-backed Python application with add/list behavior, two tests, and existing AGENTS.md instructions. DevThread initialization preserved the instructions' SHA-256 hash and added missing state.
Understanding recorded the existing behavior and the missing completion operation. Mapping created two requirements and tasks. TASK-002 depended on the existing add/list task, TASK-001.
Reconcile and choose
Baseline tests verified adding and listing tasks. A bounded source/test search found no complete_task or completion function, so completion was assessed missing. With TASK-001 done, TASK-002 was the sole eligible ready task.
Implement and verify
An acceptance test was added before implementation. The test command failed because complete_task could not be imported. The example copy then gained that operation; the original fixture was unchanged.
| Exact command, from the example directory | Recorded outcome |
|---|---|
python -m unittest discover -s tests -v before the new acceptance test | Passed: 2 baseline tests |
python -m unittest discover -s tests -v with the new test before implementation | Failed: ImportError, completion operation absent |
python -m unittest discover -s tests -v after implementation | Passed: 4 tests |
| Concurrent-writer check | Not run: outside this example's criteria; no harness supplied |
The checks cover completing a task while preserving its ID and other tasks, handling an unknown ID without changing stored data, and retaining baseline behavior.
Saved outcome
TASK-002 moved to review. Its evidence was assessed verified for the stated checks, and its report retained the earlier failure and the not-run concurrency check. The map validated; status reported no eligible ready task. Nothing automatically set it to done.
Run the existing sample from the DevThread checkout root:
Push-Location .\examples\taskboard-session
python -m unittest discover -s tests -v
Pop-Location
python .\skill\devthread\scripts\validate.py .\examples\taskboard-session
python .\skill\devthread\scripts\status.py .\examples\taskboard-session --jsonThe new-project and existing-application process scenarios are also exercised by the root test suite. Fresh Codex chats separately checked skill discovery, understand, and recovery with status and next. These are scoped local observations, as described in compatibility.