7-csk-deploy
Skill1 starsChanged 4 months agoMITView on GitHub
Release, package, deploy, or hand off an Approved feature using the architecture-approved target. Use only after recorded QA evidence and branch integration, with a readiness review before external action and a final review of the actual result before Deployed bookkeeping or tagging.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- 1-csk-initSkill
- 2-csk-write-specSkill
- 3-csk-architectureSkill
- 4-csk-frontendSkill
- 5-csk-backendSkill
- 6-csk-qaSkill
- audit-plan-loopSkill
- claude-skill-proxy-syncSkill
- create-worker-worktreeSkill
- csk-adopt-plan-scaffoldSkill
- csk-helpSkill
- csk-refineSkill
- csk-startSkill
- finish-branchSkill
- pull-main-ffSkill
- review-loopSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
18 rules. About 0 are common, 9 are distinctive, and 9 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
9
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
When this skill starts, first give the user one or two plain-language sentences in the user's language: which step this is, what happens …
- Very specific: line 29
Run /csk-start and read .claude/rules/workflow-state.md.
- Distinctive: line 31
Read architecture, master feature, engineering principles, loop policy, feature index, owning feature spec, QA results, and release …
- Very specific: line 42
If any gate is missing, stop and route to the owning skill.
- Very specific: line 45
If deploySkill is configured, complete readiness checks here, then hand release execution to that named repository-local skill.
- Distinctive: line 61
Run project-native build, test, lint, type, package, security, artifact, and smoke checks required by the release.
- Very specific: line 67
Run /review-loop on the release plan and pre-release evidence before any publish, upload, tag, store, cloud, production, or other external …
- Distinctive: line 93
After every external boundary, inspect actual state.
- Distinctive: line 99
Never rerun a possibly non-idempotent release blindly.
- Distinctive: line 104
Verify that the released artifact installs, starts, loads, or runs on its actual target.
- Distinctive: line 107
Check health, logs, errors, permissions, configuration, migrations, external integrations, and monitoring appropriate to the target.
- Very specific: line 113
If release or smoke verification fails, keep feature status Approved, record the observed external state and recovery evidence, and do not …
- Very specific: line 119
After the release attempt and smoke checks, run /review-loop again on the actual artifact/environment and captured evidence.
- Very specific: line 130
Do not update Deployed, create a final release tag, or close release bookkeeping while a Critical/High result finding remains unresolved.
- Distinctive: line 135
Only after final review passes:
- Distinctive: line 137
If the approved release process requires a Git tag, create and push it only with the explicit tag authority obtained for the release plan, …
- Very specific: line 144
Update the feature spec and feature index together to Deployed.
- Very specific: line 158
Before ending, preserve only durable unfinished or blocked work through /csk-start (show proposed rows as PENDING-AUTH and ask once).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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 / Packaging / Deployment
Release only what QA approved, through the target and process recorded in the architecture. A plan, successful local build, or user request to "deploy" does not by itself prove release readiness or authorize external actions.
Step Banner
When this skill starts, first give the user one or two plain-language sentences in the user's language: which step this is, what happens now, why it comes now (one benefit or avoided risk of doing it at this point), what comes next, and which user approval this step will ask for before anything is saved, committed, or released. Keep it to at most three sentences; do not expand it into a tutorial.
Beispiel: "Schritt 7 von 7 — Release: Wir veröffentlichen jetzt nur, was die Qualitätsprüfung bestanden hat, und sichern jeden Schritt mit Belegen und einem Rückweg ab. Danach ist das Feature ausgeliefert."
Entry contract
- Run
/csk-startand read.claude/rules/workflow-state.md. - Resolve repository-local adapter config and source-of-truth paths.
- Read architecture, master feature, engineering principles, loop policy, feature index, owning feature spec, QA results, and release instructions.
- Require authoritative feature status
Approvedplus recorded QA evidence and no unresolved Critical/High finding. - Require an architecture-approved release/package/deploy target and rollback, recovery, or rerun path.
- Inspect live Git state. If verified feature work remains on a non-default
branch, run
/finish-branchbefore release. Do not release from an unmerged or ambiguous branch unless the documented release process explicitly requires it and the user approves that exception.
If any gate is missing, stop and route to the owning skill. Do not use a quick-fix
waiver to bypass Approved status or evidence.
If deploySkill is configured, complete readiness checks here, then hand release
execution to that named repository-local skill. Resume this skill only with
direct release-result evidence.
Phase 1: Harden and verify the release plan
Apply the planning hardening policy for non-trivial releases. Verify:
- target and version fit the architecture and approved feature scope;
- exact build/package/deploy commands and artifact inputs;
- local/offline, online/cloud, privacy, data-egress, licensing, and secret rules;
- configuration, migration, compatibility, platform, performance, cost, and resource assumptions;
- backup, rollback, recovery, rerun, monitoring, and handoff paths;
- how partial success can be detected and recovered without duplicate effects.
Run project-native build, test, lint, type, package, security, artifact, and smoke checks required by the release. Verify that environment examples contain dummy values only and that no secret or local-only artifact is included.
Phase 2: Readiness review before external action
Run /review-loop on the release plan and pre-release evidence before any
publish, upload, tag, store, cloud, production, or other external mutation.
Disposition every Critical/High finding. The review must cover target fit, QA
evidence, artifact contents, privacy/security, rollback, observability, and
partial-failure recovery.
Present the user with:
- exact target, version, artifact, and commands/actions;
- evidence and remaining risks;
- external systems affected;
- rollback/recovery path;
- whether a Git tag or release record is proposed.
Obtain explicit approval for the exact external release actions. Treat tagging, pushing a tag, creating a hosted release, publishing an artifact, and deploying as distinct actions unless the user explicitly approves the bundled plan.
Phase 3: Execute and observe
- Build or select the exact approved artifact and record its identity/checksum where the architecture calls for it.
- Execute only approved release actions through the approved target or delegated skill.
- Capture command output, job/build IDs, artifact/version identifiers, URLs, and timestamps as available.
- After every external boundary, inspect actual state. Never infer completion from a submitted request alone.
- On failure or timeout, stop automatic progression, determine whether any part
succeeded, and preserve an exact recovery task through
/csk-startwhen the issue will survive the session.
Never rerun a possibly non-idempotent release blindly. A partial upload, package publication, tag, hosted release, or deployment is incomplete until reconciled.
Phase 4: Post-release verification
- Verify that the released artifact installs, starts, loads, or runs on its actual target.
- Smoke-test the released feature and critical affected journeys.
- Check health, logs, errors, permissions, configuration, migrations, external integrations, and monitoring appropriate to the target.
- Confirm no secrets, local-only data, prompts, caches, raw model output, temporary files, or sensitive logs escaped into the artifact or environment.
- Exercise or prove the rollback/recovery/rerun path to the documented degree.
If release or smoke verification fails, keep feature status Approved, record
the observed external state and recovery evidence, and do not tag or claim
Deployed.
Phase 5: Final review of the actual result
After the release attempt and smoke checks, run /review-loop again on the
actual artifact/environment and captured evidence. This is separate from the
readiness review. Confirm:
- intended target and version are really live or handed off;
- release evidence is direct and internally consistent;
- QA assumptions still hold in the target environment;
- security, privacy, processing-mode, performance, and observability promises remain true;
- rollback and partial-state risks are resolved or explicitly block completion.
Do not update Deployed, create a final release tag, or close release
bookkeeping while a Critical/High result finding remains unresolved.
Phase 6: Bookkeeping and optional tag
Only after final review passes:
- If the approved release process requires a Git tag, create and push it only
with the explicit tag authority obtained for the release plan, then verify
both local and remote tag targets. A required-tag failure keeps the feature
Approvedand creates one exact recovery task; do not advance status. - Record release details, date, target, version, artifact/URL, commit SHA, required tag, verification, rollback path, and both review results in the feature spec.
- Update the feature spec and feature index together to
Deployed; reread both. - Close the owning durable task through
/csk-startonly when its acceptance condition is met and typed evidence exists.
Release evidence may be an artifact path/hash, URL, version, verified tag,
commit, job ID, or handoff record. Missing evidence leaves status Approved.
Completion report
Report readiness review, approved external actions, actual target/version,
artifact and commit identity, smoke/monitoring result, final review, rollback
state, feature/task bookkeeping, tag state, and remaining risks. Say
incomplete for partial external state.
Before ending, preserve only durable unfinished or blocked work through
/csk-start (show proposed rows as PENDING-AUTH and ask once). If a release
branch remains open after a verified release result, route it to
/finish-branch.
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.

