Omh Release Cut

Release Cut

This is an OMH release-cut workflow skill, projected for Agent Skills hosts (Claude Code, Codex, Cursor, opencode, OpenClaw, pi).

Why This Exists

release-cut exists because nothing owned deciding a release: deploy-and-monitor watches a rollout that is already happening, and a cut, a canary, or a rollback decision reached an image card or a watch lane with no rollback trigger at all.

First Steps

• State what goes in, what is held back, and the version before proposing the cut.
• Name the rollback trigger and its exact command before the first stage ships.

Do Not Use When

• The rollout is already running and the ask is watching its health signals and post-deploy status; use omh-deploy-and-monitor.
• Production is down or degraded and the work is commanding the incident; use omh-live-incident-response.
• The ask is a readiness audit across observability, security, and operations before launch; use omh-production-audit.
• The ask is drafting the commit message or the pull request description for a change; use omh-commit-pr-authoring.
• The app ships through the App Store or Google Play -- signing, a TestFlight or Play testing-track beta, a phased release, a store hotfix; use omh-mobile-release.

Examples

Good example:

• Prompt: cut a release and tag it
• Expected behavior: List what goes in and what is held, derive the version, and prepare release_plan/v1: freeze, bump every version surface, tag, run the release workflow, pass its approval, publish, and curate notes, with rollback_trigger/v1 naming the signal, the threshold, and the exact command.
• Why: A release whose rollback is decided during the outage is a release with no rollback.

Bad example:

• Prompt: ship it, we will figure out rollback if something breaks
• Expected behavior: Keep the plan not ready, name the missing rollback trigger and command, and ask who runs it.
• Why: The rollback decision made under pressure is the one most likely to be wrong.

Completion Checklist

• The contents, the held items, and the version are stated.
• Every rollout stage has a traffic share, a bake time, and a promotion criterion.
• The rollback trigger names a signal, a threshold, and the exact command.
• The readiness verdict is ready only when the trigger and the command are named.
• Nothing was tagged, published, deployed, or rolled back by OMH.

Recovery Notes

• If the release mechanism is unknown, ask for the workflow or command and every version surface it bumps before proposing a cut.
• If there is no traffic split, stage by environment or by cohort and say that the canary is coarse.

When to Use

Use when a release is being decided or undone: what goes into it, its version, how it is tagged and published, how it rolls out in stages behind a canary, or rolling back a deploy that already shipped. The output is a release plan whose rollback has a named trigger and the exact command that performs it; OMH prepares, and the host or CI executes.

Copy code
Strong routing signals: `release-cut`, `release cut`, `cut a release`, `cut the release`, `cut a new release`, `tag a release`, `tag the release`, `release candidate`, `what goes in the release`, `version the release`, `semver bump`, `roll back the last deploy`, `roll back the deploy`, `roll back the release`, `roll back to the previous version`, `rollback trigger`, `rollback command`, `canary`, `canary release`, `canary deploy`, `set up a canary`, `staged rollout`, `progressive rollout`, `percentage rollout`

Catalog Metadata

Category: planning Phase: release-cut Quality tier: rollback-trigger-gated Reasoning demand: standard

Quality bar:

• Derive the version from the change classes, not from a feeling about size.
• Load references/release-and-rollback-method.md for the cut sequence, the rollout stage table, and rollback by mechanism instead of recalling them.
• Give every rollout stage a promotion criterion read from a signal, not a clock alone.
• Name the rollback trigger as a signal, a threshold, and a window, and the command as the literal command line.
• Keep prepared, dispatched, observed, and published as separate states for every step.

Required inputs:

• the changes since the last release, and the current version and versioning scheme
• how a release is cut here: the workflow or command, the approval gates, and every version surface it bumps
• the deploy target, the traffic split available, and the health signals with their normal range
• the rollback mechanism: redeploy of the previous artifact, a flag, or a revert, and who may run it
• observed evidence for any published, promoted, or rolled-back claim

Expected outputs:

• release_scope/v1
• release_plan/v1
• rollout_stages/v1 when the release is staged
• rollback_trigger/v1
• release_readiness_verdict/v1

Artifact expectations:

• release_scope/v1 lists what goes in and what is held back, and derives the version from the change classes: a breaking change, a feature, or a fix
• release_plan/v1 orders the cut: freeze the release branch, bump every version surface, tag, run the release workflow, pass its approval gate, publish, and curate the notes
• rollout_stages/v1 gives each stage its traffic share, its bake time, and the promotion criterion read from a named health signal
• rollback_trigger/v1 names the signal and threshold that trigger the rollback, the exact command that performs it, and who runs it
• release_readiness_verdict/v1 reads ready only when a named rollback trigger and its exact command are stated; otherwise it names what is missing

Safety rules:

• A release plan cannot be ready without a named rollback trigger and the exact command that performs it; release_readiness_verdict/v1 names what is missing instead.
• Decide the rollback before the release: the trigger, the threshold, the command, and the person who runs it are written down before the first stage ships.
• Nothing merges to the release branch between starting a cut and its tag push; a merge in that window breaks the atomic push of the bump and the tag.
• OMH never tags, publishes, deploys, approves, or rolls back; a prepared plan is never a cut release or a performed rollback.
• A rollback that crosses a database migration or a published package names what cannot be undone and how it is contained.

Runtime Evidence

Use the current host's own tools and subagent/task mechanism when available; otherwise run the same lanes sequentially or name the unavailable capability. A prepared plan, handoff, checklist, or skill installation is not execution, review, CI, merge-readiness, or merge evidence. Record actual tool results, or not_observed / not_available, in the record; never invent dispatch or host accounting. Treat supplied context as advisory, not proof of hidden memory reads or writes. State scope, constraints, verification, and the stop condition before work. Reply in the user's own words and the host's own voice: the host's persona owns reply language, tone, speech level, and sentence endings, progress updates included (where it sets no language, use the one the user wrote in), and OMH shapes structure and content only; OMH's record terms (surface, lane, wrapper, handoff, evidence boundary, not_observed) stay in records and tool calls, never in the sentence the user reads unless they ask about one; and when a real unresolved user decision or a useful briefing discussion remains, offer one specific, content-linked question. Completed work needs no question; omit unrelated non-action footers and generic offers to continue. Supporting paths are relative to this skill directory; sibling skill paths are relative to its parent. Resolve them from the host-provided skill base directory ({baseDir} on hosts that provide it), never a hardcoded install location. A named workflow not installed here is unavailable, not permission to emulate its host-specific capabilities. Verify through the real surface before done.