Correct four task records that said shipped work was not done - #217
Merged
Conversation
`tasks.md` decides whether a change may be archived, and four of them were wrong in the same direction: - `run-with-harness-prefills-the-run` recorded 0 of 16 done, having shipped entirely in #202. - `usage-from-acp`, `event-guard-covers-every-kind` and `usage-visible-while-running` each left their "run the checks" and "add a changeset" items open, with the checks run and the changesets already released in 0.50.0. One cause, not four slips. Tasks were ticked, the files copied into a worktree and committed, and only then were the checks run and `npx changeset` invoked — after the commit that would have recorded them. The same order produced the same omission four times, always at the last two items of a list. Every tick here was verified against `main` rather than recalled: the symbol the task names existing in the file it names, the test the task names existing by its own title, and for a changeset item the released entry in the relevant CHANGELOG. Fifteen of sixteen items in the prefill change verified this way; the sixteenth is human-only and stays open. The diff was then re-read item by item against that list to confirm nothing else was ticked — 21 boxes, matching 15 + 2 + 2 + 2 exactly. Every human-only item stays open, including `ci-audit-own-job` 3.6, where half is observed (the audit is now its own check) and half is not (that the merge gate reports independently *when the audit fails*, which needs a failure to demonstrate). Closing it on the observed half is the inference this change exists to refuse. `openspec/README.md` gains the order that avoids this — run, then tick, then commit — along with what the failure looks like from outside, since that is how it will be recognised next time: a change whose work is plainly in `main`, whose `tasks.md` reads as untouched, and which the archive step then refuses. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
tasks.mddecides whether a change may be archived, and four of them were wrong in the same direction.run-with-harness-prefills-the-runinitialCommandKindis inAiPanel.tsx, changeset released in webui 1.25.0usage-from-acpevent-guard-covers-every-kindusage-visible-while-runningOne cause, not four slips
Tasks were ticked, the files copied into a worktree and committed, and only then were the checks run and
npx changesetinvoked — after the commit that would have recorded them. The same order produced the same omission four times, always at the last two items of a list:run the checksandadd a changeset.Verified, not recalled
Every tick was checked against
main: the symbol the task names existing in the file it names, the test the task names existing by its own title, and for a changeset item the released entry in the relevantCHANGELOG.md. Fifteen of sixteen items in the prefill change verified this way; the sixteenth is human-only.The diff was then re-read item by item against that list to confirm nothing else was ticked — 21 boxes, matching 15 + 2 + 2 + 2 exactly. Correcting a wrong record by assumption would just make it wrong in the other direction, and harder to notice.
Human-only items all stay open
Including
ci-audit-own-job3.6, where half is observed (the audit is now its own check) and half is not — that the merge gate reports independently when the audit fails needs a failure to demonstrate. Closing it on the observed half is the inference this change exists to refuse.Stopping it recurring
openspec/README.mdgains the order — run, then tick, then commit — and what the failure looks like from outside, since that is how it will be recognised: a change whose work is plainly inmain, whosetasks.mdreads as untouched, and which the archive step then refuses.🤖 Generated with Claude Code