Release

---
name: release
description: Verify, write the CHANGELOG, cut a version — then hand the release sequence back
allowed-tools: Read, Grep, Glob, Edit, Write, Agent, Bash(git status:*), Bash(git diff:*), Bash(git log:*), Bash(git show:*), Bash(git fetch:*), Bash(git add:*), Bash(git commit:*), Bash(git rev-parse:*), Bash(git merge-base:*), Bash(npm:*), Bash(pnpm:*), Bash(yarn:*), Bash(pytest:*), Bash(python:*), Bash(go:*), Bash(cargo:*), Bash(make:*)
disable-model-invocation: true
---
Release **preparation** orchestrator for the **current branch**. It runs its own
short mechanical checks, writes the CHANGELOG, bumps the version and
commits — then **stops and reports**. It never pushes, opens a PR, merges,
tags, or publishes: those are yours to authorize by name. The general docs
sweep runs earlier, in `/branch-review`.

It does not review code. Review is a separate command that must have run
first.

## Guardrails
- **Spawn a worker, mid tier stated explicitly** (omitted inherits the
  parent's, not the balanced one; not cheapest/fastest either — judgment
  degrades there; never a vendor model name). Fall back to running inline if
  your tool has no subagent mechanism.
- **Escalate, never assume.** Anything you cannot decide, cannot verify, or
  that this spec does not cover → **stop and report it to the orchestrator**
  (the main session). Never improvise, never widen scope, never fix a finding
  you noticed along the way.
- **Nothing leaves the machine.** No `git push`, no `gh`, no `npm publish`,
  under any circumstance — not even if every gate is green. You report the
  sequence; a human authorizes it.

## Phase 0 — Preflight (current branch, always)
- **Release the branch you are on.** No branch argument, no branch creation.
- **On `main` → stop and ask** what should be released. `main` is only ever
  the merge target; never release it, never commit to it.
- **The tree must be clean — do not commit for the user.** Run `git status
  --porcelain`. Any line at all — modified, staged, or untracked — is a
  **stop**: list the paths and say "commit this to the branch, re-run
  `/branch-review`, then re-run `/release`." Committing here would create a
  commit *after* the review and so fail Phase 0.5 by its own rule on the very
  next step; a phase that guarantees the next phase fails is not a phase.
- `git fetch origin`; the release diff is `origin/main...HEAD`. Empty →
  **stop**, nothing to release.
- **Record both version numbers.** Read the **local** version (`package.json`
  or this project's equivalent) and, if the project has a publish path, the
  **published** one (`npm view <pkg> version`, or the registry equivalent).
  Report them side by side. Local *ahead* of published means a version was cut
  on a branch and never published — Phase 3 cannot see that gap unless you
  record it here, and a worker that cannot see it will re-cut a number that
  already exists.
- Print a one-line plan: branch · commit count · files changed · HEAD SHA ·
  local version → published version.

## Phase 0.5 — Review precondition (do not skip)
A review must have run on this branch **at the current HEAD SHA**.

**Compare the SHAs yourself; do not settle for an answer.** Run `git rev-parse
HEAD` and compare it against the `sha:` line (the one starting exactly
`sha:` — never `self-review-sha:`, a separate bookmark `/self-review` owns and
`/branch-review` only carries forward) in `.amp/remember/last-review.md`,
which `/branch-review` writes. Asking the
orchestrator "did a review run?" puts the question to the one party with an
incentive to say yes, so its word is not evidence — and neither is a SHA
quoted from a chat message, which is the same claim in another costume and is
gone after a compaction or a handover. Read the file; match the two strings.
**No such file, or no `sha:` line in it = no review**, never a pass. A review
that predates this file's introduction has no record, so it does not count.

- **No review**, or no recorded SHA obtainable → **stop**: "No review at
  `<sha>`. Run `/branch-review medium` (or `/code-review medium`) first."
- **Stale** — recorded SHA ≠ `git rev-parse HEAD` → **stop** and ask for a
  re-review, **unless every file**
  in `git diff --name-only <recorded sha>..HEAD` is forgiven. **This overrides
  that:** a merge or rebase of `origin/main` after the review is never forgiven,
  even when it brings only docs. A file is forgiven
  if it's under `docs/`, a `*.md`
  at the repo root, **or** on the record's `docs:` line — Stage 4 legitimately
  writes docs outside `docs/`/root too (`packages/subagentic-manual.md`,
  `packages/claude/AGENT.md`, and siblings), and its own commit is what put
  them on that line. Run this and read what it prints — every printed path
  must also be on `docs:`, or it's **stale**, stop, re-review; no output means
  every file was under `docs/` or root already, also not stale:
  ```
  git diff --name-only <recorded sha>..HEAD | grep -vE '^(docs/|[^/]+\.md$)'
  ```
  Do not take "it's just docs" on trust. This is what makes "all findings
  fixed" checkable instead of promised.
  **No exceptions beyond the three forgiven shapes above.** The fix ledger is
  normally gitignored, so appending to it moves nothing and this never comes
  up. A repo that tracks `.amp/` instead will see a ledger commit land
  after the review and make it stale — `fix-ledger.md` is neither under
  `docs/`/root nor ever on `docs:` (`/branch-review` only appends to or prunes it, it
  never sweeps it). That is the rule working, not a case to carve out:
  re-review, or leave the ledger uncommitted until after the release.
- **`tests:` line** — tests exit 0 **and** build exit 0 or `N/A: <reason>`
  **and** recorded `sha:` = HEAD → covered, do not re-run. Otherwise (HEAD
  moved past `sha:` by forgiven commits, line absent, build part missing, or
  `NOT RUN`) → run the project's real test command, and build if there is
  one, yourself in Phase 1 and cite command + exit code. Any non-zero exit →
  **stop**.
- **`coverage:` naming any stage `NOT RUN`** → **stop**. A `ready` from a run
  that skipped the security stage is not the same fact as one that did not,
  and this line is the only place the difference is visible to you.
- **Stage-2 and sweep evidence** — run this one line and read its exit code;
  nonzero → **stop**: "review record lacks stage-2 evidence — re-run
  `/branch-review`". It needs all 11 `s2 <key>:` lines (each `ran:` or `N/A:`,
  never `NOT RUN` or an unfilled template) and a `sweep:` line that is not
  `deferred`. A record from before these lines existed fails it — intended.
  The keys are listed literally rather than derived: the security spec is not
  reliably readable from here, and a test pins this list to security's.
  ```
  f=.amp/remember/last-review.md; ok=1; for k in secrets tenant-isolation rate-limiting error-handling authorization data-access injection auth-session trust-boundaries config dependencies; do grep -qE "^s2 ${k}: (ran|N/A):" "$f" || ok=0; done; [ "$ok" = 1 ] && grep -qE '^sweep: (ran|main):' "$f"
  ```
- **`verdict: blocked` in the record** → **stop**, even when the SHA matches.
  Read that line as mechanically as the `sha:` one. A matching SHA proves a
  review ran here; it says nothing about what the review concluded, and
  leaving the conclusion to the orchestrator's recollection restores exactly
  the unverified claim this file replaced. Only `verdict: ready` with a
  matching SHA is a pass.
- **Reviewed at this SHA with findings outstanding** → **stop**. Findings are
  resolved before a release is cut.

This phase runs **before** `/release` writes anything, so the CHANGELOG-and-
bump commit it makes later cannot invalidate the review it just checked.

Report this block, every field filled from what you ran (never "all checks
passed"):
```
sha: <recorded> vs <HEAD> match yes|no · verdict: <value> · coverage: <line> · s2-check: exit <code> · tests: covered | re-run <cmd> exit <code> · stale-grep: <output | empty>
```
`stale-grep` is the output of the `git diff --name-only <sha>..HEAD | grep -vE …`
check above, or `empty` when it printed nothing.

This is the only thing guaranteeing the branch was reviewed *and* security
scanned, so treat a missing answer as a **stop**, never as a pass.

## Phase 1 — Verify
Detect the stack first (`package.json`, `pyproject.toml`, `go.mod`,
`Cargo.toml`, `Makefile`) and run only what exists. Report each item **pass /
fail / N/A** with the exact command and exit code; a check not run is a
**fail**; N/A needs a stated reason. Read the exit code off the bare command
(`cmd > /tmp/out 2>&1; e=$?`), never a pipeline — `$?` after a pipe is the
last element's status. A test command that can outlast your tool's default
command timeout must be run with a longer timeout (or in the background and
waited on until it exits); a run that timed out is not a pass, and the report
cites the suite's totals as well as the exit code. Emit a coverage row: `ran? ✓/✗` · evidence · verdict.
A ✗ is **Blocked 🛑**.

- **Lint / format clean** — only if a linter or formatter is configured.
- **Migrations ready** — only if the project has a schema / migrations: they
  apply cleanly and are ordered.
- **In sync with `origin`** — not behind `origin/main`; a never-pushed branch
  passes.
- **Tests / build** — only per the `tests:` bullet in Phase 0.5 (HEAD moved
  past `sha:`, line absent, build part missing, or `NOT RUN`).

Security is **not** re-run here — it is stage 2 of the review, already
confirmed in Phase 0.5.

## 🚦 Gate
- **Any Critical** (failing tests, broken build) → **stop**, report, escalate.
- **Warnings, or anything you cannot confidently decide** → **stop**,
  summarize, escalate. Do not weigh it yourself.
- **All clean** → continue.

## Phase 2 — CHANGELOG
The general docs sweep (guide/context doc, PRD, README, findings/learnings)
happens in `/branch-review` (Stage 4), already covered by Phase 0.5. This
phase only writes `CHANGELOG.md`: an entry covering every user-visible
change in `origin/main..HEAD`, read from the commit bodies (not just
subjects), under the headings the file already uses. If the file already has
a `## [Unreleased]` section, that is this release's draft: check it against
the commits, add anything missing, and retitle it `## [X.Y.Z] - YYYY-MM-DD`.
Otherwise write a new entry under that title. Never leave both. An Added
entry means at least a minor bump. Past entries are history — leave them.

**Locate the heading safely — the bare string is not a safe anchor.** A
CHANGELOG that documents release tooling can quote `## [Unreleased]` inside a
shipped release note's own prose, so a blind find-and-replace on that string
can rewrite history instead of the draft. Anchor on the full heading with its
surrounding newlines (`"\n## [Unreleased]\n"`) — only the real heading has
both. Count matches before writing: exactly one is required; zero or more
than one is a **stop** — report it, do not guess which one is real. After
writing, verify every other occurrence of the bare string is byte-identical
to before, and that the previous version's heading and `---` separators
survived.

**This is the worker's job, start to finish.** The orchestrator checks the
entry; it does not redo or patch it.

## Phase 3 — Cut (local only)
1. **Version bump** — pick the semver level from the change (patch / minor /
   major; **ask if ambiguous**) and update `package.json`. The local-vs-
   published gap you recorded in Phase 0 **is** an ambiguity: if local is ahead
   of published, a version was cut and never published, so ask whether to
   publish that number or bump past it. Never silently re-cut a version that
   already exists locally. The bump must land
   on the branch, before any merge — a version committed to `main` directly,
   or added after the merge, breaks the tag/package match.
2. **Commit** — `release: vX.Y.Z — <summary>`, including the CHANGELOG and
   the bump.

Then **stop.** Nothing else.

This release commit is the one commit `/release` itself adds after the
review. It touches `package.json` for the version bump — not under `docs/`
or root, and never on a `docs:` line, since `/release` isn't `/branch-review`'s
Stage 4 sweep — so it moves HEAD past the reviewed SHA with a file Phase
0.5's rule doesn't forgive, even though the same commit's `CHANGELOG.md`
would be. Running `/release` twice on the same branch without a re-review
therefore still correctly stops as stale, as it always did.

## Report — the sequence, for a human to authorize
Print the evidence, then hand back the exact remaining steps so the
orchestrator can run them on the user's named go:

> **Cut ✅ vX.Y.Z on `<branch>`** — checks green, CHANGELOG updated, release
> commit made locally. Reviewed at `<sha>`.
> Ready when you are:
> 1. `git push -u origin <branch>`
> 2. `gh pr create` into `main`
> 3. `gh pr checks <pr> --watch` — **merge only on green.** Every gate before
>    this one ran on the same machine; CI is the only differently-configured
>    instrument in the chain, and this is the first time it sees the branch.
>    A test that passes locally because of a path, a fixture, or a tool that
>    exists only on your box fails here and nowhere earlier. Read the exit
>    code off the bare command. Red → stop, fix, re-review, and start again.
> 4. `gh pr merge --admin --merge --delete-branch` (main is PR-protected;
>    owner-authorized admin merge on a solo repo). **Keep `--merge`** — `gh`
>    requires an explicit merge-method flag (`--squash` / `--merge` /
>    `--rebase`), and a release PR lands as a merge commit so the branch's own
>    commits stay on main.
> 5. `git tag vX.Y.Z` on `main` and push the tag
> 6. Publish **if this project has a publish path** (e.g.
>    `gh workflow run publish.yml`) — manual by design
> 7. Verify it is actually live (`npm view <pkg> version`, and the published
>    tarball's contents), not the working tree

**Every exit code in this sequence is read off the bare command, including
the ones you type yourself.** The rule is not just for the worker: a
pipeline reports its last element's status, so `gh run watch --exit-status |
tail -2; echo $?` prints `0` for a failed run. That has already turned a red
CI into a green reading in a real release.

Final line: **Cut ✅ (vX.Y.Z — ready to push)** or **Blocked 🛑** with the
specific reason.