The walk back from a merged PR. Routed to by /ca:cleanup.
The work has landed on the default branch. What is left is local: the branch you\nare standing on, and whatever the run left in the tree. This skill returns you to\na clean, fast-forwarded default checkout without losing anything you have not\nexplicitly agreed to lose.
\nIt is ordinary lifecycle work and never needs /ca:override. When it\ncannot proceed, it names the gate that stopped it. Issue #308 recorded the\nalternative: with no owner for this transition, the routing loop reached for\n/ca:chore, then for /ca:override — a bypass manufactured to cover a\ncoverage hole.
Read this, or STOP and surface the gap — never guess the default-branch name:
\n${CLAUDE_PROJECT_DIR}/.codearbiter/CONTEXT.md — the default-branch name.If HEAD is already the default branch, there is no transition to make. Report the\ndirty state read-only and exit; branch pruning across other branches belongs to\n/ca:standup.
A branch that looks merged is not merged. Establish it against the network, not\nagainst a stale local ref. This repo squash-merges by default, so SHA-ancestry\nalone is the wrong instrument for most landings: it holds for a fast-forward or\na merge-commit landing, but a squash merge writes a new commit with no SHA\nlineage back to the branch, and --is-ancestor will report non-zero for a\nbranch that landed cleanly. The gate is content-containment, proven by\nwhichever instrument fits the landing, always reported as a fact:
git fetch the remote holding the default branch. A fetch that fails STOPs —\nan unfetched comparison proves nothing.git merge-base --is-ancestor HEAD origin/<default>. Holds\nfor a fast-forward or merge-commit landing. If it holds, that is the\nproof: report it and move on.gh pr list --head \"$(git branch --show-current)\" --state merged --json number,state,headRefOid,mergeCommit.\nA MERGED PR whose headRefOid equals local git rev-parse HEAD proves\nevery commit on this branch rode that PR's squash into the default\nbranch. Report the PR number, that headRefOid == HEAD, the merge\ncommit (mergeCommit.oid in that JSON), and the ancestry check's\nnegative result — all as facts, not as a failure. Corroborate with\ngit diff --quiet origin/<default> HEAD when it happens to hold; a\nnon-empty diff alongside a valid PR proof is normal (the default\nbranch advanced since the merge landed) and is reported as a fact, never\ntreated as a failure. The PR-record identity — MERGED plus\nheadRefOid == HEAD — is the load-bearing proof; the diff is\ncorroboration only when it happens to be fresh, never a requirement.gh is unavailable or no PR record exists —\ngit diff --quiet origin/<default> HEAD (the tree is byte-identical to\nthe fetched default) is an acceptable fallback proof. Report it as such.If none of the three hold, STOP exactly as today. Name the un-landed commits\nand route to /ca:pr; nothing is deleted here.
\n\nDeliberate deviation from issue #586's suggested contract: that suggestion\nrequired the squash-merge proof's diff to be empty. That requirement\nre-breaks this gate the moment any later PR merges to the default branch —\nthe common state, not an edge case. The PR-record identity is what proves\ncontainment; the diff is corroboration, demoted from requirement to fact.
\n
Gate: the remote is fetched and HEAD is proven contained in the fetched default\nbranch — by ancestry, by the PR-record squash proof, or by the byte-identical\nfallback — or the skill has stopped.
\nList every dirty tracked change, every untracked file, and every stash reachable\nfrom this branch. Classify each into exactly one of three, and say why for each:
\nPresent the classification before touching anything. A file whose class you\ncannot establish is reported as unique with the reason you could not classify it.
\nGate: every dirty, untracked, and stashed artifact carries a class and a stated\nreason, with unclassifiable items counted as unique.
\nPer item, in the order Phase 2 listed them. Never batch:
\n/ca:commit it on this branch before the\ntransition, move it aside, or leave it in place and stop the cleanup. It is\ndiscarded only if the user explicitly confirms that item by name, with the\nPhase 2 reasoning in view.A stash is never dropped here. Stashes are reported with git stash show as the\nsuggested next step, the same report-and-route contract /ca:standup holds.
If anything the user chose to keep would block the checkout, STOP and say so\nrather than removing it anyway — but distinguish the two causes before naming\nthe artifact as the blocker:
\n<default> makes that difference larger than reality: content\nthe kept artifact never actually touches can still collide with what a\ncurrent default branch would carry. That gap is not the user's problem and\nnot the kept artifact's fault. It is resolved by Phase 4's fast-forward-first\nstep, not by discarding anything here. Issue #586 recorded the observed\nfailure mode: a stale local main — 513 lines behind on\n.codearbiter/gate-events.log alone — made this STOP name a correctly-kept\n38-line artifact as the blocker, which invites exactly the silent discard\nthis phase exists to prevent.Gate: every item is resolved by an explicit per-item decision, and the working\ntree is clean enough to check out the default branch — or the skill has stopped\nwith the blocker named, correctly attributed to a genuine conflict or a stale\nlocal ref.
\nOnly after Phase 3 leaves the tree safe:
\ngit fetch origin <default>:<default>. Git refuses a non-fast-forward\nupdate of a ref that is not currently checked out, so the --ff-only\nguarantee is preserved by the instrument itself — this cannot silently\nrewrite the local default ref, only advance it or refuse. This is what\nkeeps a stale local default from blocking (or worse, misattributing) the\ncheckout below. This step can itself be refused for the same reason a\ncheckout can be — <default> checked out in another worktree — in which\ncase report that as the (correctly attributed) blocker, per Phase 3, and\nstop; do not treat the refusal as license to skip ahead.git branch --show-current). A checkout silently fails when another worktree\nholds the branch, and every step after this one would otherwise run against\nthe wrong branch.--ff-only (belt over the braces of step 1 — a no-op when\nthe pre-checkout fetch already brought the local ref current). A divergence\nmeans the default branch moved in a way this skill will not reconcile:\nreport it and stop. Never a merge commit, never a rebase, never a reset.Gate: the local default ref was fast-forwarded before the checkout — or the\nfast-forward's refusal was itself reported as a fact and the skill stopped —\nHEAD is confirmed on the default branch by re-read, and the belt-and-braces\n--ff-only pull either succeeded (typically a no-op) or was reported as a\nrefused divergence.
Offer deletion of the now-merged local branch, with the Phase 1 containment\nproof restated. One confirmation, naming the branch.
\ngit branch -d first, always. Never reach for -D on the assumption that a\nrefusal means -d can't handle a squash merge — it often can. git branch -d's safety check accepts a branch merged into its upstream, not only\ninto HEAD: with an upstream still configured and equal to the tip, -d\nsucceeds (with a warning) even for a squash merge that --is-ancestor alone\nwould call unmerged. Do not conclude a refusal here means -D is required —\ncheck the reason first.-D path. -d typically refuses once the remote branch\nhas been auto-deleted and pruned, because there is then no upstream left for\n-d to test reachability against. When that happens, git branch -D is\npermitted, but ONLY when both hold:\nMERGED and headRefOid == HEAD) — not a bare ancestry pass, not a\nstale or assumed proof.-d's safety check tests SHA-reachability, and the\nsquash proof has already shown that instrument is the wrong one for this\nrepo's merge mode. The PR-record proof plus a named confirmation replaces\nthe check that -d can no longer run — it does not bypass it. Everywhere\nelse, -D stays forbidden exactly as before: if -d refuses and Phase 1's\nproof was anything other than the PR-record squash proof, report both and\nstop.Declining leaves the branch in place. That is a normal outcome, not a failure.
\nGate: the branch is deleted only after an explicit confirmation naming it, via\n-d, or via -D restricted to the sanctioned path above with the proof\nrestated, with the remote untouched.
One short summary: what landed, what was removed, what was kept, and where HEAD\nis now. State declines as declines — a cleanup the user stopped halfway is a\ncorrect outcome reported plainly, not an error.
\nHEAD is contained in the fetched default branch\nbefore any deletion — via ancestry, via the PR-record squash proof (MERGED\nand headRefOid == HEAD), or, without gh, via a byte-identical tree diff.\nMUST NOT infer merge state from a : gone] upstream alone.git fetch origin <default>:<default>)\nbefore checking it out, or report the fast-forward's own refusal (e.g.\n<default> checked out in another worktree) as the blocker and stop. MUST\nre-read the current branch after checkout before acting on it.--ff-only, and MUST NOT merge, rebase, or reset to reach the default\nbranch.git branch -d, and MAY use -D ONLY when the Phase-1 squash-merge\nproof held this run and -d refused, with that proof restated and the branch\nnamed in the confirmation. Everywhere else -D is forbidden. MUST NOT delete a\nremote branch, force-push, or write to the default branch./ca:standup does./ca:override when blocked. Name the gate instead.