name	release-deploy
description	Use when user mentions 'deploy version', 'release v1.0', 'go-live', 'tag release', 'production cutover', 'release checklist', 'deploy plan', or before tagging any production release. Guides through `release-deploy-standard.md` per-bump-type checklist + go-live runbook + smoke test + rollback procedure.
user-invocable	true

Skill: Release Deploy Guide

Version: 1.0.0 Created: 2026-05-06 Pairs with rule: .claude/rules/release-deploy-standard.md

When to Use

User mentions: "deploy version X", "release v1.0", "go-live", "tag release", "production cutover", "release checklist", "deploy plan ready?"
Before tagging any git tag matching v[0-9]+.[0-9]+.[0-9]+*
After receiving production incident requiring hotfix release
Quarterly release audit cycle

Trigger conditions

User: "Let's deploy v1.0.0"
User: "Tag release v0.9.0-beta"
User: "Ready to go-live?"
User: "Run release checklist for v1.0.0"
User: "Production deploy plan?"
User: "Hotfix release v1.0.1"

What this skill does

Identify release type (PRE-RELEASE / PATCH / MINOR / MAJOR) from version tag
Load applicable per-bump-type checklist from rule §3
Check each artifact status (✅ done / ⏳ pending / ❌ missing)
For pending/missing: link to gap or propose new gap
Run pre-deploy verification commands
Generate deploy command sequence (Helm + Terraform + kubectl)
Run smoke test post-deploy
Capture deploy outcome → wave-history.jsonl + GitHub Release

Methodology — 7 stages

Stage 1: Identify release type

Parse version tag per versioning-policy.md §3:

v0.9.0-beta       → PRE-RELEASE (beta)
v1.0.0-rc.1       → PRE-RELEASE (rc)
v1.0.0            → MAJOR + first PRODUCTION
v1.0.1            → PATCH
v1.1.0            → MINOR
v2.0.0            → MAJOR (persona expansion)

Stage 2: Load applicable checklist

From .claude/rules/release-deploy-standard.md §3:

§3.1 PRE-RELEASE — subset
§3.2 PATCH — subset (extends §3.1)
§3.3 MINOR — most items (extends §3.2)
§3.4 MAJOR + first PROD — full

Stage 3: Status check each artifact

For each item in checklist:

# Examples
[deploy plan]       → check parent doc link in PR description
[smoke test]        → check scripts/smoke-test.sh exists
[rollback]          → check documents/05-guides/operations/runbooks/rollback-runbook.md
[secrets manager]   → check AWS Secrets Manager configured
[DNS]               → dig kitehub.vn +short
[CDN]               → curl -I https://kitehub.vn | grep CF-Ray
[email]             → send test email
[health check]      → curl <url>/actuator/health
[smoke test]        → ./scripts/smoke-test.sh

Mark each: ✅ done / ⏳ pending (link gap) / ❌ missing (file new gap)

Stage 4: Pre-deploy verification

Run automated checks:

CI green on tag commit
Database backup taken
Quality audit /100 ≥ threshold (PRE-RELEASE 80; PROD MAJOR 85)
Staging deploy successful (if MAJOR)
Migration plan documented (if schema change)

Stage 5: Deploy execution

Per parent deploy plan steps. Human-in-the-loop required for critical commands (per release-deploy-standard.md §9):

# Agent CAN do (prep/observation):
- Generate deploy command sequence
- Suggest Helm values
- Parse logs
- Run smoke tests

# Agent SHOULD NOT do (production execution):
- terraform apply
- helm upgrade --install (production namespace)
- kubectl apply (production)
- DNS cutover
- Database migrations on prod

User executes commands manually OR via separately-authenticated CI/CD workflow.

Stage 6: Smoke test post-deploy

./scripts/smoke-test.sh <prod-url>
# Expect: all 15+ assertions pass

# If fail: trigger rollback procedure (per GAP-378)

Stage 7: Capture outcome

Append wave-history.jsonl entry per feedback_wave_history_append_required.md
Update documents/03-planning/changelogs/CHANGELOG-vX.Y.Z.md
Create GitHub Release với changelog body
Update status page (resolved if pre-existing maintenance)

Skill output format

# Release Deploy Status — vX.Y.Z

**Release type:** PRE-RELEASE / PATCH / MINOR / MAJOR
**Target environment:** beta / staging / production
**Standard reference:** `.claude/rules/release-deploy-standard.md` §3.X

## Artifact checklist (N/M done)

### Operational Excellence
- [✅] <item> — <reference>
- [⏳] <item> — pending GAP-XXX
- [❌] <item> — MISSING; recommend file new gap

### Security
- [✅] ...
- ...

### Reliability
- ...

## Pre-deploy verification

- [✅/❌] CI green on tag
- [✅/❌] Backup taken
- [✅/❌] Quality audit /100 ≥ threshold
- ...

## Deploy plan

Per parent doc: `documents/03-planning/roadmap/release-1-deploy-plan.md` §X

Commands ready for execution (human-in-the-loop):
1. ...
2. ...

## Smoke test

Run: `./scripts/smoke-test.sh <url>`
Expected: 15+ assertions pass

## Rollback ready?

Per: `documents/05-guides/operations/runbooks/rollback-runbook.md`

## Decision

🟢 GO — All blocking items satisfied
🟡 GO WITH CAVEATS — N items pending; override trailer required citing follow-up gaps
🔴 NO-GO — N blocking items missing; close gaps before retry

Skill contents

This SKILL.md — methodology
reference/per-bump-checklist.md — detailed checklist tables (mirror rule §3)
reference/standards-cheatsheet.md — Well-Architected/Twelve-Factor/DORA quick refs
scripts/check-release-readiness.sh — automated artifact check script (defer follow-up)

Gotchas

Don't run terraform apply or helm upgrade on production from agent — per release-deploy-standard.md §9 + output-review-mandate.md Section 6, production execution = human-required
Smoke test failure → trigger rollback procedure, not "ignore and proceed"
Override trailer must reference follow-up gap — bare override blocked per rule §5
Skip if user just asks "what's deploy status" without intent to actually deploy — reply with current state, don't run full skill
Beta release ≠ production release — different checklists; verify version tag pattern carefully
DNS cutover irreversible without prep — prep rollback DNS records BEFORE cutover

Rule: .claude/rules/release-deploy-standard.md — the standard
Rule: .claude/rules/output-review-mandate.md §3 — production deployment review row
Plan: documents/03-planning/roadmap/release-1-deploy-plan.md — Release 1 specific instance
Plan: documents/03-planning/roadmap/versioning-policy.md — version convention
ADR: documents/02-architecture/adr/ADR-015-aws-agent-plugins-evaluation.md — agent role evaluation
GAP: GAP-381 — Claude agent deploy framework evaluation
Gaps: GAP-369..380 — Release 1 deploy artifact gaps

Log

2026-05-06 (v1.0.0): Skill created paired same-PR with release-deploy-standard.md per rule-change-process.md §6.5 Enforcement Parity Mandate. Triggered by user feedback to codify deploy standard via rule + skill + workflow thay vì ad-hoc gap files. Standards grounded: AWS Well-Architected + Twelve-Factor + DORA + OWASP. Self-test on Release 1 v1.0.0 (rule §11) verifies skill activates correctly + maps existing gap state.

name	release-deploy
description	Use when user mentions 'deploy version', 'release v1.0', 'go-live', 'tag release', 'production cutover', 'release checklist', 'deploy plan', or before tagging any production release. Guides through `release-deploy-standard.md` per-bump-type checklist + go-live runbook + smoke test + rollback procedure.
user-invocable	true

Skill: Release Deploy Guide

Version: 1.0.0 Created: 2026-05-06 Pairs with rule: .claude/rules/release-deploy-standard.md

When to Use

User mentions: "deploy version X", "release v1.0", "go-live", "tag release", "production cutover", "release checklist", "deploy plan ready?"
Before tagging any git tag matching v[0-9]+.[0-9]+.[0-9]+*
After receiving production incident requiring hotfix release
Quarterly release audit cycle

Trigger conditions

User: "Let's deploy v1.0.0"
User: "Tag release v0.9.0-beta"
User: "Ready to go-live?"
User: "Run release checklist for v1.0.0"
User: "Production deploy plan?"
User: "Hotfix release v1.0.1"

What this skill does

Identify release type (PRE-RELEASE / PATCH / MINOR / MAJOR) from version tag
Load applicable per-bump-type checklist from rule §3
Check each artifact status (✅ done / ⏳ pending / ❌ missing)
For pending/missing: link to gap or propose new gap
Run pre-deploy verification commands
Generate deploy command sequence (Helm + Terraform + kubectl)
Run smoke test post-deploy
Capture deploy outcome → wave-history.jsonl + GitHub Release

Methodology — 7 stages

Stage 1: Identify release type

Parse version tag per versioning-policy.md §3:

v0.9.0-beta       → PRE-RELEASE (beta)
v1.0.0-rc.1       → PRE-RELEASE (rc)
v1.0.0            → MAJOR + first PRODUCTION
v1.0.1            → PATCH
v1.1.0            → MINOR
v2.0.0            → MAJOR (persona expansion)

Stage 2: Load applicable checklist

From .claude/rules/release-deploy-standard.md §3:

§3.1 PRE-RELEASE — subset
§3.2 PATCH — subset (extends §3.1)
§3.3 MINOR — most items (extends §3.2)
§3.4 MAJOR + first PROD — full

Stage 3: Status check each artifact

For each item in checklist:

# Examples
[deploy plan]       → check parent doc link in PR description
[smoke test]        → check scripts/smoke-test.sh exists
[rollback]          → check documents/05-guides/operations/runbooks/rollback-runbook.md
[secrets manager]   → check AWS Secrets Manager configured
[DNS]               → dig kitehub.vn +short
[CDN]               → curl -I https://kitehub.vn | grep CF-Ray
[email]             → send test email
[health check]      → curl <url>/actuator/health
[smoke test]        → ./scripts/smoke-test.sh

Mark each: ✅ done / ⏳ pending (link gap) / ❌ missing (file new gap)

Stage 4: Pre-deploy verification

Run automated checks:

CI green on tag commit
Database backup taken
Quality audit /100 ≥ threshold (PRE-RELEASE 80; PROD MAJOR 85)
Staging deploy successful (if MAJOR)
Migration plan documented (if schema change)

Stage 5: Deploy execution

Per parent deploy plan steps. Human-in-the-loop required for critical commands (per release-deploy-standard.md §9):

# Agent CAN do (prep/observation):
- Generate deploy command sequence
- Suggest Helm values
- Parse logs
- Run smoke tests

# Agent SHOULD NOT do (production execution):
- terraform apply
- helm upgrade --install (production namespace)
- kubectl apply (production)
- DNS cutover
- Database migrations on prod

User executes commands manually OR via separately-authenticated CI/CD workflow.

Stage 6: Smoke test post-deploy

./scripts/smoke-test.sh <prod-url>
# Expect: all 15+ assertions pass

# If fail: trigger rollback procedure (per GAP-378)

Stage 7: Capture outcome

Append wave-history.jsonl entry per feedback_wave_history_append_required.md
Update documents/03-planning/changelogs/CHANGELOG-vX.Y.Z.md
Create GitHub Release với changelog body
Update status page (resolved if pre-existing maintenance)

Skill output format

# Release Deploy Status — vX.Y.Z

**Release type:** PRE-RELEASE / PATCH / MINOR / MAJOR
**Target environment:** beta / staging / production
**Standard reference:** `.claude/rules/release-deploy-standard.md` §3.X

## Artifact checklist (N/M done)

### Operational Excellence
- [✅] <item> — <reference>
- [⏳] <item> — pending GAP-XXX
- [❌] <item> — MISSING; recommend file new gap

### Security
- [✅] ...
- ...

### Reliability
- ...

## Pre-deploy verification

- [✅/❌] CI green on tag
- [✅/❌] Backup taken
- [✅/❌] Quality audit /100 ≥ threshold
- ...

## Deploy plan

Per parent doc: `documents/03-planning/roadmap/release-1-deploy-plan.md` §X

Commands ready for execution (human-in-the-loop):
1. ...
2. ...

## Smoke test

Run: `./scripts/smoke-test.sh <url>`
Expected: 15+ assertions pass

## Rollback ready?

Per: `documents/05-guides/operations/runbooks/rollback-runbook.md`

## Decision

🟢 GO — All blocking items satisfied
🟡 GO WITH CAVEATS — N items pending; override trailer required citing follow-up gaps
🔴 NO-GO — N blocking items missing; close gaps before retry

Skill contents

This SKILL.md — methodology
reference/per-bump-checklist.md — detailed checklist tables (mirror rule §3)
reference/standards-cheatsheet.md — Well-Architected/Twelve-Factor/DORA quick refs
scripts/check-release-readiness.sh — automated artifact check script (defer follow-up)

Gotchas

Don't run terraform apply or helm upgrade on production from agent — per release-deploy-standard.md §9 + output-review-mandate.md Section 6, production execution = human-required
Smoke test failure → trigger rollback procedure, not "ignore and proceed"
Override trailer must reference follow-up gap — bare override blocked per rule §5
Skip if user just asks "what's deploy status" without intent to actually deploy — reply with current state, don't run full skill
Beta release ≠ production release — different checklists; verify version tag pattern carefully
DNS cutover irreversible without prep — prep rollback DNS records BEFORE cutover

Rule: .claude/rules/release-deploy-standard.md — the standard
Rule: .claude/rules/output-review-mandate.md §3 — production deployment review row
Plan: documents/03-planning/roadmap/release-1-deploy-plan.md — Release 1 specific instance
Plan: documents/03-planning/roadmap/versioning-policy.md — version convention
ADR: documents/02-architecture/adr/ADR-015-aws-agent-plugins-evaluation.md — agent role evaluation
GAP: GAP-381 — Claude agent deploy framework evaluation
Gaps: GAP-369..380 — Release 1 deploy artifact gaps

Log

2026-05-06 (v1.0.0): Skill created paired same-PR with release-deploy-standard.md per rule-change-process.md §6.5 Enforcement Parity Mandate. Triggered by user feedback to codify deploy standard via rule + skill + workflow thay vì ad-hoc gap files. Standards grounded: AWS Well-Architected + Twelve-Factor + DORA + OWASP. Self-test on Release 1 v1.0.0 (rule §11) verifies skill activates correctly + maps existing gap state.

release-deploy

Skill: Release Deploy Guide

When to Use

Trigger conditions

What this skill does

Methodology — 7 stages

Stage 1: Identify release type

Stage 2: Load applicable checklist

Stage 3: Status check each artifact

Stage 4: Pre-deploy verification

Stage 5: Deploy execution

Stage 6: Smoke test post-deploy

Stage 7: Capture outcome

Skill output format

Skill contents

Gotchas

Related

Log

Skill: Release Deploy Guide

When to Use

Trigger conditions

What this skill does

Methodology — 7 stages

Stage 1: Identify release type

Stage 2: Load applicable checklist

Stage 3: Status check each artifact

Stage 4: Pre-deploy verification

Stage 5: Deploy execution

Stage 6: Smoke test post-deploy

Stage 7: Capture outcome

Skill output format

Skill contents

Gotchas

Related

Log