release-delivery
Skill0 starsChanged 9 days agoMITView on GitHub
Use when setting up or changing CI pipelines, releasing a project or learning its release protocol, preparing a tag or deployment, writing rollback or runbook notes, verifying a deployed revision, packaging a deployment for a client, preparing a client handover, or reviewing documentation after a release.
In this repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- backend-engineeringSkill
- change-planningSkill
- change-reviewSkill
- device-healthSkill
- genai-engineeringSkill
- instruction-maintenanceSkill
- pakemSkill
- pakem-updateSkill
- project-bootstrapSkill
- root-cause-debuggingSkill
- sensitive-operationsSkill
- state-and-handoffSkill
- ui-ux-qualitySkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Every repository has CI from the first commit.
- Distinctive: line 57
Follow release protocol: on a release, read the section and run it.
- Distinctive: line 72
Follow operational alerting for failure reports and notifications.
- Distinctive: line 82
Every handover includes the items in handover, using the runbook template and the user guide template, in the client's language.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Release and Delivery
Production actions need the owner's explicit instruction (sensitive-operations).
This skill applies in full when the project's AGENTS.md says Delivery: us. With
Delivery: others, keep CI green and stop at the merge request; follow the team's release
process and never change their pipelines, deploy files, or environments unless asked.
With Delivery: nobody, only checks and version tags apply. A missing Delivery: line
is a gap to ask, not a default.
1. CI
Every repository has CI from the first commit. Use the CI templates (CI standard; adoption guide): always format, lint, type check, tests, build, secret scan, dependency audit; plus database and deployment jobs switched on by capabilities.
2. Deployment
For servers you or your pipeline control, follow deployment: deploy from a commit only, immutable release per commit, preflight, migrate, activate with automatic restore, revision-matching health check, rollback for every environment, one deploy per environment at a time. Use the deploy templates.
The standard is deployment contracts v1 (ADR 0009) as revised by ADR 0010 (accepted 2026-10-21):
- Each project chooses its builder and trigger: its organization's CI runner, the owner's device with the host building from signed source, or free GitHub runners. Never assume a hosted CI.
- Checks run on every push and merge request on a free runner, separately from releases; a release refuses a commit whose checks did not pass (a release tag reuses its commit's).
- Nothing paid unless the project or its profile approves it (ADR 0012).
- Each of the owner's devices has one dedicated key that connects and signs releases; a shared builder (a CI runner, a client platform) gets one key per project.
- Direct delivery is
pakem release(stagestablefor services): the owner's device sends signed source and the host builds it (contracts §13). The registry mode in the contracts is valid only where its CI minutes and registry are free.
One release engine per host, signed releases, build once, health checks that prove the revision, and automatic restore apply to every builder.
For client-operated hosting, build a versioned handoff package from a tagged commit (handover). Never deploy loose files or zips of uncommitted work.
3. Releases
Each project keeps its release protocol in a ## Release section of its AGENTS.md.
Follow release protocol: on a release, read the section
and run it; without one, learn the protocol from the repository (or interview the owner
when the method is undecided) and record it with the first release. The owner's go covers
the whole protocol, production included; stop only for a merge, waiting for the owner,
and on failure.
- Production releases are tagged
vMAJOR.MINOR.PATCHonmain, with any pre-release suffix the project's scheme uses; staging deploys frommain. - Each release has notes (CHANGELOG entry or release description): changes, migrations, configuration changes, rollback notes, known limits.
- Evidence (deployed revision, health, smoke results) is tied to the exact commit and
recorded in
PROJECT_STATE.mdand the change'sverification.md.
4. Operational alerting
Follow operational alerting for failure reports and notifications.
5. After release
Review documentation against the deployed behaviour: runbooks, configuration names, commands, ownership, known constraints. Update only what changed.
6. Client handover
Every handover includes the items in handover, using the runbook template and the user guide template, in the client's language.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

