@@ -463,13 +463,15 @@ ever matters more than PR independence, re-adding the one-line CI step
 A release no longer gets a hand-written checklist. The sweep for `vN` is a **filter
 over this ledger**: `since == vN` (the new capabilities) ∪ all `P0` (the standing
 smoke) ∪ any item whose `source` cites a PR in the release. The tracking issue for the
-sweep links here and hosts discussion; results stay OUT of the repo — every run files
+sweep links here and hosts discussion; results stay OUT of the repo — every sweep run files
 one `qa-run` GitHub issue as its record (text only: the verdict table + a reproduction
 rule per failure — RUNNER.md rule 2's authentication/authorization carve-out excepted,
 and it binds the extracted card too — never screenshots; `runs/` is git-ignored), and a
 run that finds a real regression extracts it into its own standalone card at close-out
 (RUNNER.md, extraction obligation — the run issue itself stays `qa-run`-only, excluded
-from triage). Item text, fixtures learned, and new traps discovered flow **back into the
+from triage). A verification re-run — one triggered by the same PR's ledger edit of that
+item — files no record: its evidence is the PR body and the item's `history` row (RUNNER.md,
+Run records). Item text, fixtures learned, and new traps discovered flow **back into the
 ledger** as revisions — that is the accumulation the one-off checklists never had.
 
 ## Relationship to what already exists
 
@@ -457,9 +457,16 @@ contradicts it, and correct it here when it does.
 
 ## Run records — the GitHub issue is the report
 
-Every completed run — **pass or fail alike** — files **one GitHub issue** as its durable
-record, labeled `qa-run` and nothing else (extraction obligation below). **Nothing lands
-in the repo** — not the JSON, not screenshots; `runs/` is git-ignored except its README.
+Every completed sweep run (a release candidate, an area or `all` run, or a maintainer-named
+target) — **pass or fail alike** — files **one GitHub issue** as its durable record, labeled
+`qa-run` and nothing else (extraction obligation below). **A verification re-run files
+none** — a run whose trigger is the same PR's ledger edit of that item. Its per-item verdict
+table goes in the PR body and the item's `history` row carries the PR as its `ref`; a
+`partial` or `blocked` gap goes in the PR body and the item's `fixtures.knownGaps`; a
+product-defect `fail` is still extracted as its own card, pointing at the PR; a stale clause
+is fixed in the same diff (`.claude/skills/checklist-test/SKILL.md` §5, 「验证重测」).
+**Nothing lands in the repo** — not the JSON, not screenshots; `runs/` is git-ignored except
+its README.
 
 **The issue is text only.** Screenshots and DOM dumps are oracles you consult *live* to
 reach a verdict — never report artifacts. What the report carries for a defect is the
 
@@ -8,8 +8,12 @@ land in the repo (`.gitignore` here tracks only this README). The durable source
 the checklist itself under `../areas/`; a run is a snapshot that goes stale the moment
 the build moves.
 
-**Where results go instead — every completed run files one `qa-run` GitHub issue** as its
-durable, **text-only** report (pass or fail alike):
+**Where results go instead — every completed sweep run** (a release candidate, an area or
+`all` run, a maintainer-named target) **files one `qa-run` GitHub issue** as its durable,
+**text-only** report (pass or fail alike). A verification re-run — one triggered by the same
+PR's ledger edit of that item — files none: its evidence is the PR body and the item's
+`history` row, with a `partial` / `blocked` gap also in `fixtures.knownGaps`
+([../RUNNER.md](../RUNNER.md), Run records, states the exception in full). A sweep's record:
 
 - the issue **body** hosts the per-clause verdict table (pass / partial / fail / blocked)
   and the scope — the selector that chose the items + the `revision` each ran against;