release-manager · daffy0208 — Claude Code skill: what it does, how to install
release-manager turns a risky deploy into stages: internal team first, then 5% of beta users, then everyone. Comes with feature flags, canary deploys and rollback rules.
MIT 135 installs ★ 39 updated 2025-12-25
npx skills add daffy0208/ai-dev-standards --skill release-manager -a claude-codeWhat release-manager does
The skill walks a release through stages. The feature goes to the whole internal team first so critical bugs surface, on days two and three it reaches 5% of beta users, then 25% and 50%, and the flag comes off at full release. Every stage is written out in YAML, with a TypeScript flag example next to it: on for the beta-testers group, otherwise enabled by a hash of user.id.
Two deployment patterns are covered. Blue-green moves traffic to a green environment and keeps blue around for rollback; you pay for double infrastructure. Canary starts on 5% of servers with an hour of watching between steps. Checklists for before, during and after the release sit in their own section, along with metrics (error rate, p95, CPU and memory), alerts and a rollback plan.
Rollback thresholds are fixed up front: past 1% errors or a performance drop over 20% and the flag goes off. The flag example calls hashUserId, which you write yourself, or the percentage rollout has nothing to hash.
What it helps with
- roll a risky feature out to 5% of users without breaking the rest
- set up a feature flag with group and percentage targeting
- pick between blue-green and canary for your infrastructure budget
- write rollback criteria before the incident, not during it
- put together a release checklist and the metrics to watch
Example prompts for the agent
- Plan the rollout of
new-dashboardin phases and set up a feature flag for it - Write the release checklist and rollback criteria for a canary deploy
- Compare blue-green and canary for our service and say which costs more in infrastructure
How to install release-manager
1Skills CLI (Vercel)
npx skills add daffy0208/ai-dev-standards --skill release-manager -a claude-code2Manually: put the files in place
~/.claude/skills/release-manager/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/release-manager/
1Skills CLI (Vercel)
npx skills add daffy0208/ai-dev-standards --skill release-manager -a codex2Manually: put the files in place
.agents/skills/release-manager/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add daffy0208/ai-dev-standards --skill release-manager -a cursor2Manually: put the files in place
.cursor/skills/release-manager/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install daffy0208/ai-dev-standards skills/release-manager2Manually: put the files in place
.github/skills/release-manager/SKILL.md1Skills CLI (Vercel)
npx skills add daffy0208/ai-dev-standards --skill release-manager -a gemini-cli2Platform CLI
gemini skills install https://github.com/daffy0208/ai-dev-standards --path skills/release-managerManually: put the files in place
.devin/skills/release-manager/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add daffy0208/ai-dev-standards --skill release-manager -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about release-manager
How do I install release-manager in Claude Code?
Run npx skills add daffy0208/ai-dev-standards --skill release-manager -a claude-code. The skill lands at ~/.claude/skills/release-manager/SKILL.md.
What does release-manager do?
It lays out a staged release: internal team first, then beta users, then wider, with feature flags, blue-green and canary deploys, checklists and a rollback plan.
When does the skill say to turn the feature off?
Once errors pass 1% or performance drops more than 20%, it treats both as rollback triggers.
Source text
SKILL.md · MIT
---
name: release-manager
description: Ship features safely with progressive rollouts, feature flags, and canary deployments. Use when deploying risky features or need gradual rollouts.
license: Complete terms in LICENSE.txt
---
# Release Manager
Ship features safely with progressive rollouts.
## Progressive Rollout Strategy
```yaml
Phase 1 - Internal (Day 1):
- 100% to internal team
- Test thoroughly
- Fix critical bugs
Phase 2 - Beta (Day 2-3):
- 5% to beta users
- Monitor errors/performance
- Collect feedback
Phase 3 - Gradual (Day 4-7):
- 25% of users
- Watch metrics closely
- 50% of users if good
- 100% if still good
Phase 4 - Full Release:
- 100% of users
- Remove feature flag
- Announce publicly
```
## Feature Flags
```typescript
// Feature flag implementation
const featureFlags = {
newDashboard: {
enabled: true,
rollout: 0.25, // 25% of users
userGroups: ['beta-testers'], // Always on for beta
}
}
function isFeatureEnabled(feature, user) {
const flag = featureFlags[feature]
// Check user group
if (user.groups.some(g => flag.userGroups.includes(g))) {
return true
}
// Check rollout percentage
const hash = hashUserId(user.id)
return (hash % 100) < (flag.rollout * 100)
}
// Usage
{isFeatureEnabled('newDashboard', user) ? (
<NewDashboard />
) : (
<OldDashboard />
)}
```
## Deployment Strategies
### Blue-Green Deployment
```yaml
Process: 1. Deploy to "green" environment
2. Test green thoroughly
3. Switch traffic to green
4. Keep blue as rollback
Pros: Instant rollback
Cons: 2x infrastructure cost
```
### Canary Deployment
```yaml
Process: 1. Deploy to 5% of servers
2. Monitor for 1 hour
3. If good, deploy to 25%
4. Monitor for 1 hour
5. If good, deploy to 100%
Pros: Gradual, safe
Cons: Slower rollout
```
## Rollback Plan
```yaml
Criteria for Rollback:
- Error rate > 1%
- Performance degradation > 20%
- Critical bug discovered
- Negative user feedback
Rollback Process: 1. Disable feature flag immediately
2. Notify team
3. Investigate issue
4. Fix and redeploy
```
## Release Checklist
### Pre-Release
- [ ] Code reviewed
- [ ] Tests passing
- [ ] Staging tested
- [ ] Feature flag configured
- [ ] Rollback plan ready
- [ ] Monitoring alerts set
### During Release
- [ ] Deploy to 5% first
- [ ] Watch error rate
- [ ] Monitor performance
- [ ] Check user feedback
- [ ] Gradually increase
### Post-Release
- [ ] Monitor for 24 hours
- [ ] Collect feedback
- [ ] Remove feature flag
- [ ] Document learnings
## Monitoring
```yaml
Key Metrics During Release:
- Error rate
- Response time p95
- CPU/memory usage
- User-reported issues
Alerts:
- Error rate > 1% → Pause rollout
- Response time > 2s → Investigate
- Memory spike > 90% → Rollback
```
## Communication
```yaml
Internal:
- Slack announcement
- Deploy log updated
- Engineering team notified
External:
- Changelog updated
- Email to power users (if major)
- Blog post (if significant)
```
## Summary
Safe releases:
- ✅ Start small (5%)
- ✅ Monitor closely
- ✅ Rollback readily
- ✅ Feature flags everywhere
- ✅ Document process
© daffy0208 · MIT · daffy0208/ai-dev-standards
Similar skills
All skills for DevOps and infrastructure →release-manager by other authors comes first.