--- name: quick-wrap description: "Short session close: commit, handle loose ends, stop. Not workstream-complete." version: 1.0.0 allowed-tools: ["Read", "Edit", "Bash", "Grep", "Glob", "Skill"] argument-hint: "[optional context]" --- # Quick Wrap — the Light End-of-Work Close Doctrine has named this exit for months without providing it. `skills/handoff/SKILL.md` says *"Workstreams end via `/workday-complete`, `/merge-to-main`, or commit-and-stop"*, and several sibling surfaces repeat the same three-way choice. Every one of those citations pointed at a ceremony that did not exist — so "commit-and-stop" meant *improvise it*, and the improvisation was usually reaching for a heavier ceremony that fit worse. This skill is that missing artifact: the short checklist for a session whose work is finished, small, and unremarkable. **It is deliberately not a cheaper `/workstream-complete`.** The entry test below is a positive gate, not a matter of taste, precisely because a light exit that is easy to reach becomes the default — and every session that takes it is a session that wrote no completion entry and reconciled no plan. If you are reading the entry test looking for a way past it, the answer is already no — with one exception, and it is written into the test rather than reasoned around it: test 2 names a review debt, and paying that debt directly is a supported route, not an override. Every other failure routes to the named sibling. --- ## Entry test — ALL four must hold Fail any one and this is not the room. Route to the named remedy instead and do not negotiate with the test; it is cheap to discharge the debt the test names and expensive to lose the capture. **The four tests measure two independent debts.** Tests 1, 3, and 4 ask *is continuity substrate owed* — a plan reconciliation, a consumed predecessor, a successor baton. Test 2 asks *is a review owed*. A session can owe a review and owe no continuity artifact at all; that is why test 2 has its own remedy and does not route to `/workstream-complete`. Routing it there conscripted a ceremony with no plan to reconcile and no predecessor to account for in order to deliver a review. | # | Condition | If it fails | |---|---|---| | 1 | **No governing plan** drove this session's work | `/workstream-complete` — a plan needs reconciling against what shipped | | 2 | **Diff is under the review brightline** — under 500 novel LOC, under 5 commits, under 4 distinct surfaces | **Scoped review, then wrap** — the named route below. Not `/workstream-complete`. | | 3 | **No predecessor consumed** — this session picked up no baton | `/workstream-complete` — a session that consumed a predecessor owes a coverage gate | | 4 | **Work is finished** — nothing in-flight for a successor | `/handoff` — and only if context pressure is genuinely forcing the stop | ### Sizing test 2 — novel LOC, code LOC The brightline counts **novel code lines**, not gross diff lines. Two carve-outs, both explicit so they are applied deliberately rather than argued for case by case: - **Relocation does not count.** Lines moved without modification — a forward-port, a revert, a file move, un-stranding previously-reviewed work, a vendored-content move — were already reviewed where they were authored, and relocating them adds no review surface. What *is* review surface in a relocation is the **seams**: dropped content, a regression leaking across, a stale citation left pointing at the old locus. Size on the seams and scope any review to them. Modified-in-transit lines are novel and do count. - **Doc-only lines do not count.** Test 2 gates a *code* review; a decision record, forensics writeup, wiki page, or handoff body is not code-review surface. This is a ruling, not an oversight — a session that lands an investigation plus a decision record should not trip a code-review brightline on prose. Doc quality is reviewed at the distillation and docs ceremonies, not here. If applying the carve-outs is what drops you under the line, say so in the step-4 report with the gross and novel counts. A carve-out claimed silently is indistinguishable from talking yourself through the gate. ### The named route — scoped review, then wrap Test 2 failing means a review is owed, not that this ceremony is wrong. Discharge it directly: 1. **Dispatch a reviewer scoped to the actual risk surface** — the seams for a relocation, the novel hunks otherwise. `Skill(coordinator:review-code)` for a diff-shaped review; a direct `coordinator:code-reviewer` dispatch when the scope is narrow and named. 2. **Integrate the findings** via the review-integrator, per standing doctrine — findings are applied, never ratified back to the PM as a list. 3. **Then run the checklist below**, reporting the review verdict and what it found in step 4. This is the only route past test 2. It is not a lighter option — it is the same review debt paid directly instead of through a ceremony that also demanded continuity artifacts the session did not owe. **It does not unlock the other three tests:** a session that fails test 1, 3, or 4 still routes to its named sibling regardless of how its review resolved. **A lesson is not a disqualifier.** A surfaced lesson used to be test 2 and is deliberately gone: capture is already discharged by the `lesson` CLI, the backlog is deep enough that one more entry is not worth a heavier ceremony, and routing on it sent single-memo sessions to a workstream close they had no workstream for. Capture the lesson through the CLI and wrap. Two further redirects that are not size questions at all: - **Branch is ready to ship** → `/merging-to-main`. Quick-wrap ends a session, never a branch. - **End of the working day, not just this session** → `/workday-complete`. It supersedes this; never run both. **Mutual exclusion.** `/quick-wrap` and `/workstream-complete` are exclusive, the same way `/handoff` and `/workstream-complete` already are. One session, one closing ceremony — closing ceremony here means the session terminators (`ceremony-calibration.md` § Session terminators: `/workstream-complete`, `/handoff`, `/workday-complete`). If you have two workstreams and genuinely different dispositions, close each separately and name which is which. --- ## The checklist Four steps. If it is taking longer than a few minutes, the entry test was wrong. **1. Commit.** Run the safe-commit mechanism — it computes this session's own safe pathspec from its touch-list (minus anything a live peer session also claims) and commits + pushes it: ```bash "${COORDINATOR_SETTINGS_HOME:-$HOME/.coordinator-claude-settings}/bin/safe-commit-offer" ``` **Do not ask whether to commit** — by explicit PM ruling, being asked was itself the defect. Run it, then report what landed. Prefer authored grouping over the mechanical default when this session did describable work: `--message ""` for one coherent group, or `--groups-json ` for several. Any path you name that is not in the computed safe pathspec is silently dropped — the boundary is computed, never caller-widened. **2. Close the loose ends.** This is the judgment step, and it is yours — nothing computes it. Sweep what this session actually touched and finish what it left open. The recurring shapes: - A **bug or improvement queue entry** this session resolved. Closure is a move, never an in-place edit: `git mv state//.yaml archive///.yaml`, then set `status: closed`, `closed_at: `, and `closed_by:` in the archived file. An entry marked closed but still sitting in `state/` is the failure mode this shape prevents. - An **inbound cross-repo memo** this session actioned → resolve it rather than leaving it in the inbox. Inbox depth is not a backlog. - A **stale marker or tracker row** this session invalidated. - **Scratch this session authored** that nobody needs tomorrow. - A **residual this session surfaced and did not close**, recorded so far only in prose — a plan body, a commit message, your own step-4 report. Prose is not a home. Route it via `coordinator-queue-append --schema bug-backlog|debt-backlog|improvement-queue`; naming an open item in step 4 and filing nothing is dumping it, not carrying it forward. Nothing to close is a perfectly ordinary outcome — say so and move on. Do not manufacture a loose end to make the step feel earned. **3. Refresh orientation.** `regenerate-orientation-cache` — so the next session boots on what is now true rather than what was true this morning. **4. Report.** Three lines, no ceremony: what landed (with the commit SHA), what you closed, and what is deliberately still open. Then stop. --- ## Anti-scope - **Never deletes, force-pushes, or rewrites history.** No `git reset --hard`, no `git checkout --` over someone else's hunk, no branch deletion, no rebase. The only mutation this skill authorizes is a scoped forward commit and a `git mv` of a queue entry into the archive. Anything irreversible is out of scope, including behind a flag. - **Writes no continuity artifact.** No handoff, no spinoff, no completion entry. If the session needs one, the entry test already routed you elsewhere — a quick-wrap that writes a handoff is a `/handoff` wearing a disguise. - **Captures no lesson itself.** A lesson does not disqualify the session (see the entry test), but this ceremony is not where it lands — write it through the `lesson` CLI, then wrap. - **Runs no test suite.** By explicit PM ruling, the "fast" tier is neither fast nor cheap, and paying for it at every light close was the defect — `Skill(coordinator:validate)` was step 1 and is deliberately gone. Do not re-add it, and do not substitute a hand-rolled `pytest` invocation for it. Test evidence this session *already* produced is reported as-is; a suite already known red is not wrapped over — that session fails the entry test and belongs in `/workstream-complete`. The heavier ceremonies keep their validate gates; this one does not have one to skip. - **Opens no review trail of its own.** Entry test 2 bounds the diff below the scale where a review is owed; above it, the scoped-review route runs *before* this checklist and reports its verdict into step 4. The checklist itself never dispatches a reviewer, and a scoped review is never a substitute for the partitioned review a plan-driven or predecessor-consuming session owes. - **Never closes a queue entry this session did not actually resolve.** Step 2 is a sweep of your own work, not a backlog-grooming pass. Closing on a hunch is worse than leaving it open. - **Does not supersede the PM's closure authority.** Authority to close a *workstream* belongs to the PM, signalled by their invoking a closing ceremony (a session terminator — see `ceremony-calibration.md` § Session terminators). This skill closes a *session* — reporting state honestly and stopping. It does not declare a workstream complete, and it never emits a "Session Complete" header, which preempts that authority and tends to coincide with leaving real follow-ups unfinished. ## Discovery Surfaced at `/workday-start` alongside the other closing ceremonies, and named in `/handoff`'s Step 0 NO-tests as the concrete destination for the "shipped, not handed-off" redirect — the citation that previously dead-ended on "commit-and-stop" with nowhere to go. ## Test surface No runtime test for the body — a skill body is prose-doctrine, not executable code, the same rationale `skills/plan/SKILL.md` § Test Surface records for its own. The applicable automated checks are skill-body lint and frontmatter validation. The entry test's four conditions are the reviewable surface; they are stated as a table specifically so drift in them is greppable. ## See also - `skills/workstream-complete/SKILL.md` — the heavy close, for an entry-test failure on tests 1, 3, or 4. Test 2 takes the scoped-review route instead. - `skills/handoff/SKILL.md` — in-flight work under context pressure; its Step 0 redirects here. - `commands/workday-complete.md` — the day-level close, which supersedes this one.