lsm / neokai / 26801282614

82%

push

github

feat(forge): backfill task.result from reportedSummary on terminal transitions (#2051)

* feat(forge): backfill task.result from reportedSummary on terminal transitions

SpaceTaskManager.setTaskStatus now backfills result from reportedSummary
when transitioning to done/blocked without an explicit result, preventing
tasks from reaching terminal status with a null result when a summary
exists.

Episode judge prompt also falls back to reportedSummary when result is
empty, so historical evidence with null result still surfaces useful
context to the Forge episode judge.

Includes unit tests for backfill behavior and prompt fallback.

* fix: backfill result from incoming reportedSummary on terminal transitions

Reviewer catch: when setTaskStatus receives options.reportedSummary but
task.reportedSummary is null, the backfill previously skipped. Now uses
options.reportedSummary ?? task.reportedSummary as the fallback source.

* fix: address review feedback on reportedSummary handling

- Pass reportedSummary through spaceTask.update RPC handler into
  setTaskStatus so backfill sees incoming summaries.
- Clear reportedSummary on reactivation (done/blocked/cancelled →
  open/in_progress, in_progress → open) to prevent stale reuse.
- Scope episode judge prompt fallback to terminal statuses only
  (done/blocked), avoiding misrepresentation of in_progress/review
  progress notes as final results.

* fix: preserve explicit null reportedSummary during terminal transitions

Reviewer catch: when spaceTask.update passes reportedSummary: null,
the ?? undefined coalescing dropped the explicit clear. Now uses
Object.hasOwn to preserve null intent. setTaskStatus accepts
reportedSummary?: string | null and only backfills when the caller
did not explicitly clear (options.reportedSummary !== null).

* fix: honor explicit null result and clear summary on review rejection

- RPC handler preserves explicit result: null using Object.hasOwn
  instead of ?? undefined.
- setTaskStatu... (continued)

9896 of 14446 branches covered (68.5%)

Branch coverage included in aggregate %.

22 of 28 new or added lines in 3 files covered. (78.57%)

82987 of 97642 relevant lines covered (84.99%)

297.14 hits per line