Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions create-dev-loop.md
Original file line number Diff line number Diff line change
Expand Up @@ -319,6 +319,8 @@ Count the soft ceiling against **non-test net LOC**. If non-test changes exceed

**Before pushing, verify each `Closes #N`.** For every issue number you plan to reference, run `gh issue view <N>` and confirm the title and body match what this PR does. Numbers carried forward from earlier session context or from summarized prior cycles are a common source of wrong-issue auto-closes — if a referenced issue describes unrelated work, omit the `Closes` reference and either file a new issue or note `No tracking issue — gap found during triage.` in the PR body.

**Then check the other direction: every closing keyword in the body must be meant.** GitHub's closing-keyword parser does not understand negation, so `Does not close #42`, `This doesn't fix #17`, and `not resolved by this PR: resolves #9` all still close the issue on merge. The issue then drops out of the next cycle's triage while the PR body reads as though leaving it open was deliberate. The keywords are `close`/`closes`/`closed`, `fix`/`fixes`/`fixed`, and `resolve`/`resolves`/`resolved`, matched case-insensitively and with or without a colon. Write one directly before an issue reference only when this PR genuinely closes that issue. To mention an issue without closing it, use the bare reference (`#42`) or a non-keyword phrase such as "tracked separately in #42" or "out of scope for this PR: #42". Before `gh pr create`, and again after any later edit to the body, read every keyword-plus-reference pair in the body and confirm each one is an intended close.

\`\`\`bash
git push -u origin {{BRANCH_PREFIX}}/<short-description>
# compose the body with the Write tool to a uniquely-named scratch file first (Phase 3 scratch-file rule)
Expand Down Expand Up @@ -370,6 +372,7 @@ Perform a self-review. This step is anchored on external signals ({{EXTERNAL_SIG
- **Docs:** every row in the Phase 7 documentation sources-of-truth table reflects the new behavior. **Perform Phase 7's table pass now, here, before scoring this item** — it is the only way the score is grounded in a check rather than in judgment, which the rubric's own framing forbids. Fix any row that is wrong, commit it onto this PR, and score from the completed pass; record in the self-review comment which rows were checked and which were corrected. Phase 7 then reduces to a re-check of whatever Phases 5–6 changed.
- **Visuals:** if the diff changes anything a README image depicts (UI layout, colours, rendering, on-screen or terminal output), the affected images are regenerated from the project's own current output in this PR — list the README's image references (`grep -Eo '!\[[^]]*\]\([^)]+\)|<img [^>]*src="[^"]+"' README.md`) and confirm each depicted surface the diff touches has its file in `git diff --name-only origin/{{DEFAULT_BRANCH}}...HEAD`. Never substitute a mockup; look at every regenerated image before committing (no secrets, emails, real user/player names, IPs). Score N/A when the diff changes nothing visible. A project with visible output but no README visual at all is a gap to file as an issue, not a FAIL on this PR.
- **Issue resolution:** every `Closes #N` issue's named surface area is actually changed; no issue is partially resolved while claiming closure.
- **Closing keywords:** every closing keyword followed by an issue reference in the PR body (`gh pr view <number> --json body`) names an issue this PR fully resolves, and no negated phrase ("does not close #N", "doesn't fix #N") appears. Quote each keyword-plus-reference pair found in the self-review.
- **{{EXTERNAL_SIGNAL_LABEL}}:** the external anchor is green on the PR head (re-confirms step 1).

**Tests-fix fallback ladder when the local anchor cannot run.** The revert-and-run experiment above needs a live local anchor. CI on the fixed tree alone cannot substitute for it — CI only ever runs the *fixed* state and can never reproduce the reverted half. Enter this ladder **only** when the tool/interpreter itself is unavailable or broken (the UNVERIFIED case above), so that no local technique can execute. A failure of one git mechanism is not that condition: if `git stash` refuses (a clean tree, a dirty tree, submodule state) but `git checkout` works — or the reverse — that *is* the experiment above via a different mechanism, so score Tests-fix from it directly and do not enter the ladder.
Expand Down
Loading