20 KiB
Human Checkpoints in Plans
Plans execute autonomously. Checkpoints formalize the interaction points where human verification or decisions are needed.
Core principle: Claude automates everything with CLI/API. Checkpoints are for verification and decisions, not manual work.
Checkpoint Types
1. checkpoint:human-verify (Most Common)
When: Claude completed automated work, human confirms it works correctly.
Use for:
- Visual UI checks (layout, styling, responsiveness)
- Interactive flows (click through wizard, test user flows)
- Functional verification (feature works as expected)
- Audio/video playback quality
- Animation smoothness
- Accessibility testing
Structure:
<task type="checkpoint:human-verify" gate="blocking">
<what-built>[What Claude automated and deployed/built]</what-built>
<how-to-verify>
[Exact steps to test - URLs, commands, expected behavior]
</how-to-verify>
<resume-signal>[How to continue - "approved", "yes", or describe issues]</resume-signal>
</task>
Key elements:
<what-built>: What Claude automated (deployed, built, configured)<how-to-verify>: Exact steps to confirm it works (numbered, specific)<resume-signal>: Clear indication of how to continue
Example: Vercel Deployment
<task type="auto">
<name>Deploy to Vercel</name>
<files>.vercel/, vercel.json</files>
<action>Run `vercel --yes` to create project and deploy. Capture deployment URL from output.</action>
<verify>vercel ls shows deployment, curl {url} returns 200</verify>
<done>App deployed, URL captured</done>
</task>
<task type="checkpoint:human-verify" gate="blocking">
<what-built>Deployed to Vercel at https://myapp-abc123.vercel.app</what-built>
<how-to-verify>
Visit https://myapp-abc123.vercel.app and confirm:
- Homepage loads without errors
- Login form is visible
- No console errors in browser DevTools
</how-to-verify>
<resume-signal>Type "approved" to continue, or describe issues to fix</resume-signal>
</task>
Example: UI Component
<task type="auto">
<name>Build responsive dashboard layout</name>
<files>src/components/Dashboard.tsx, src/app/dashboard/page.tsx</files>
<action>Create dashboard with sidebar, header, and content area. Use Tailwind responsive classes for mobile.</action>
<verify>npm run build succeeds, no TypeScript errors</verify>
<done>Dashboard component builds without errors</done>
</task>
<task type="checkpoint:human-verify" gate="blocking">
<what-built>Responsive dashboard layout at /dashboard</what-built>
<how-to-verify>
1. Run: npm run dev
2. Visit: http://localhost:3000/dashboard
3. Desktop (>1024px): Verify sidebar left, content right, header top
4. Tablet (768px): Verify sidebar collapses to hamburger
5. Mobile (375px): Verify single column, bottom nav
6. Check: No layout shift, no horizontal scroll
</how-to-verify>
<resume-signal>Type "approved" or describe layout issues</resume-signal>
</task>
Example: Xcode Build
<task type="auto">
<name>Build macOS app with Xcode</name>
<files>App.xcodeproj, Sources/</files>
<action>Run `xcodebuild -project App.xcodeproj -scheme App build`. Check for compilation errors in output.</action>
<verify>Build output contains "BUILD SUCCEEDED", no errors</verify>
<done>App builds successfully</done>
</task>
<task type="checkpoint:human-verify" gate="blocking">
<what-built>Built macOS app at DerivedData/Build/Products/Debug/App.app</what-built>
<how-to-verify>
Open App.app and test:
- App launches without crashes
- Menu bar icon appears
- Preferences window opens correctly
- No visual glitches or layout issues
</how-to-verify>
<resume-signal>Type "approved" or describe issues</resume-signal>
</task>
2. checkpoint:decision
When: Human must make choice that affects implementation direction.
Use for:
- Technology selection (which auth provider, which database)
- Architecture decisions (monorepo vs separate repos)
- Design choices (color scheme, layout approach)
- Feature prioritization (which variant to build)
- Data model decisions (schema structure)
Structure:
<task type="checkpoint:decision" gate="blocking">
<decision>[What's being decided]</decision>
<context>[Why this decision matters]</context>
<options>
<option id="option-a">
<name>[Option name]</name>
<pros>[Benefits]</pros>
<cons>[Tradeoffs]</cons>
</option>
<option id="option-b">
<name>[Option name]</name>
<pros>[Benefits]</pros>
<cons>[Tradeoffs]</cons>
</option>
</options>
<resume-signal>[How to indicate choice]</resume-signal>
</task>
Key elements:
<decision>: What's being decided<context>: Why this matters<options>: Each option with balanced pros/cons (not prescriptive)<resume-signal>: How to indicate choice
Example: Auth Provider Selection
<task type="checkpoint:decision" gate="blocking">
<decision>Select authentication provider</decision>
<context>
Need user authentication for the app. Three solid options with different tradeoffs.
</context>
<options>
<option id="supabase">
<name>Supabase Auth</name>
<pros>Built-in with Supabase DB we're using, generous free tier, row-level security integration</pros>
<cons>Less customizable UI, tied to Supabase ecosystem</cons>
</option>
<option id="clerk">
<name>Clerk</name>
<pros>Beautiful pre-built UI, best developer experience, excellent docs</pros>
<cons>Paid after 10k MAU, vendor lock-in</cons>
</option>
<option id="nextauth">
<name>NextAuth.js</name>
<pros>Free, self-hosted, maximum control, widely adopted</pros>
<cons>More setup work, you manage security updates, UI is DIY</cons>
</option>
</options>
<resume-signal>Select: supabase, clerk, or nextauth</resume-signal>
</task>
3. checkpoint:human-action (Rare)
When: Action has NO CLI/API and requires human-only interaction, OR Claude hit an authentication gate during automation.
Use ONLY for:
- Authentication gates - Claude tried to use CLI/API but needs credentials to continue (this is NOT a failure)
- Email verification links (account creation requires clicking email)
- SMS 2FA codes (phone verification)
- Manual account approvals (platform requires human review before API access)
- Credit card 3D Secure flows (web-based payment authorization)
- OAuth app approvals (some platforms require web-based approval)
Do NOT use for pre-planned manual work:
- Manually deploying to Vercel (use
vercelCLI - auth gate if needed) - Manually creating Stripe webhooks (use Stripe API - auth gate if needed)
- Manually creating databases (use provider CLI - auth gate if needed)
- Running builds/tests manually (use Bash tool)
- Creating files manually (use Write tool)
Structure:
<task type="checkpoint:human-action" gate="blocking">
<action>[What human must do - Claude already did everything automatable]</action>
<instructions>
[What Claude already automated]
[The ONE thing requiring human action]
</instructions>
<verification>[What Claude can check afterward]</verification>
<resume-signal>[How to continue]</resume-signal>
</task>
Key principle: Claude automates EVERYTHING possible first, only asks human for the truly unavoidable manual step.
Example: Email Verification
<task type="auto">
<name>Create SendGrid account via API</name>
<action>Use SendGrid API to create subuser account with provided email. Request verification email.</action>
<verify>API returns 201, account created</verify>
<done>Account created, verification email sent</done>
</task>
<task type="checkpoint:human-action" gate="blocking">
<action>Complete email verification for SendGrid account</action>
<instructions>
I created the account and requested verification email.
Check your inbox for SendGrid verification link and click it.
</instructions>
<verification>SendGrid API key works: curl test succeeds</verification>
<resume-signal>Type "done" when email verified</resume-signal>
</task>
Example: Credit Card 3D Secure
<task type="auto">
<name>Create Stripe payment intent</name>
<action>Use Stripe API to create payment intent for $99. Generate checkout URL.</action>
<verify>Stripe API returns payment intent ID and URL</verify>
<done>Payment intent created</done>
</task>
<task type="checkpoint:human-action" gate="blocking">
<action>Complete 3D Secure authentication</action>
<instructions>
I created the payment intent: https://checkout.stripe.com/pay/cs_test_abc123
Visit that URL and complete the 3D Secure verification flow with your test card.
</instructions>
<verification>Stripe webhook receives payment_intent.succeeded event</verification>
<resume-signal>Type "done" when payment completes</resume-signal>
</task>
Example: Authentication Gate (Dynamic Checkpoint)
<task type="auto">
<name>Deploy to Vercel</name>
<files>.vercel/, vercel.json</files>
<action>Run `vercel --yes` to deploy</action>
<verify>vercel ls shows deployment, curl returns 200</verify>
</task>
<!-- If vercel returns "Error: Not authenticated", Claude creates checkpoint on the fly -->
<task type="checkpoint:human-action" gate="blocking">
<action>Authenticate Vercel CLI so I can continue deployment</action>
<instructions>
I tried to deploy but got authentication error.
Run: vercel login
This will open your browser - complete the authentication flow.
</instructions>
<verification>vercel whoami returns your account email</verification>
<resume-signal>Type "done" when authenticated</resume-signal>
</task>
<!-- After authentication, Claude retries the deployment -->
<task type="auto">
<name>Retry Vercel deployment</name>
<action>Run `vercel --yes` (now authenticated)</action>
<verify>vercel ls shows deployment, curl returns 200</verify>
</task>
Key distinction: Authentication gates are created dynamically when Claude encounters auth errors during automation. They're NOT pre-planned - Claude tries to automate first, only asks for credentials when blocked.
See references/cli-automation.md "Authentication Gates" section for more examples and full protocol.
Execution Protocol
When Claude encounters type="checkpoint:*":
- Stop immediately - do not proceed to next task
- Display checkpoint clearly:
════════════════════════════════════════
CHECKPOINT: [Type]
════════════════════════════════════════
Task [X] of [Y]: [Name]
[Display checkpoint-specific content]
[Resume signal instruction]
════════════════════════════════════════
- Wait for user response - do not hallucinate completion
- Verify if possible - check files, run tests, whatever is specified
- Resume execution - continue to next task only after confirmation
For checkpoint:human-verify:
════════════════════════════════════════
CHECKPOINT: Verification Required
════════════════════════════════════════
Task 5 of 8: Responsive dashboard layout
I built: Responsive dashboard at /dashboard
How to verify:
1. Run: npm run dev
2. Visit: http://localhost:3000/dashboard
3. Test: Resize browser window to mobile/tablet/desktop
4. Confirm: No layout shift, proper responsive behavior
Type "approved" to continue, or describe issues.
════════════════════════════════════════
For checkpoint:decision:
════════════════════════════════════════
CHECKPOINT: Decision Required
════════════════════════════════════════
Task 2 of 6: Select authentication provider
Decision: Which auth provider should we use?
Context: Need user authentication. Three options with different tradeoffs.
Options:
1. supabase - Built-in with our DB, free tier
2. clerk - Best DX, paid after 10k users
3. nextauth - Self-hosted, maximum control
Select: supabase, clerk, or nextauth
════════════════════════════════════════
Writing Good Checkpoints
DO:
- Automate everything with CLI/API before checkpoint
- Be specific: "Visit https://myapp.vercel.app" not "check deployment"
- Number verification steps: easier to follow
- State expected outcomes: "You should see X"
- Provide context: why this checkpoint exists
- Make verification executable: clear, testable steps
DON'T:
- Ask human to do work Claude can automate (deploy, create resources, run builds)
- Assume knowledge: "Configure the usual settings" ❌
- Skip steps: "Set up database" ❌ (too vague)
- Mix multiple verifications in one checkpoint (split them)
- Make verification impossible (Claude can't check visual appearance without user confirmation)
When to Use Checkpoints
Use checkpoint:human-verify for:
- Visual verification (UI, layouts, animations)
- Interactive testing (click flows, user journeys)
- Quality checks (audio/video playback, animation smoothness)
- Confirming deployed apps are accessible
Use checkpoint:decision for:
- Technology selection (auth providers, databases, frameworks)
- Architecture choices (monorepo, deployment strategy)
- Design decisions (color schemes, layout approaches)
- Feature prioritization
Use checkpoint:human-action for:
- Email verification links (no API)
- SMS 2FA codes (no API)
- Manual approvals with no automation
- 3D Secure payment flows
Don't use checkpoints for:
- Things Claude can verify programmatically (tests pass, build succeeds)
- File operations (Claude can read files to verify)
- Code correctness (use tests and static analysis)
- Anything automatable via CLI/API
Checkpoint Placement
Place checkpoints:
- After automation completes - not before Claude does the work
- After UI buildout - before declaring phase complete
- Before dependent work - decisions before implementation
- At integration points - after configuring external services
Bad placement:
- Before Claude automates (asking human to do automatable work) ❌
- Too frequent (every other task is a checkpoint) ❌
- Too late (checkpoint is last task, but earlier tasks needed its result) ❌
Complete Examples
Example 1: Deployment Flow (Correct)
<!-- Claude automates everything -->
<task type="auto">
<name>Deploy to Vercel</name>
<files>.vercel/, vercel.json, package.json</files>
<action>
1. Run `vercel --yes` to create project and deploy
2. Capture deployment URL from output
3. Set environment variables with `vercel env add`
4. Trigger production deployment with `vercel --prod`
</action>
<verify>
- vercel ls shows deployment
- curl {url} returns 200
- Environment variables set correctly
</verify>
<done>App deployed to production, URL captured</done>
</task>
<!-- Human verifies visual/functional correctness -->
<task type="checkpoint:human-verify" gate="blocking">
<what-built>Deployed to https://myapp.vercel.app</what-built>
<how-to-verify>
Visit https://myapp.vercel.app and confirm:
- Homepage loads correctly
- All images/assets load
- Navigation works
- No console errors
</how-to-verify>
<resume-signal>Type "approved" or describe issues</resume-signal>
</task>
Example 2: Database Setup (Correct)
<!-- Claude automates everything -->
<task type="auto">
<name>Create Upstash Redis database</name>
<files>.env</files>
<action>
1. Run `upstash redis create myapp-cache --region us-east-1`
2. Capture connection URL from output
3. Write to .env: UPSTASH_REDIS_URL={url}
4. Verify connection with test command
</action>
<verify>
- upstash redis list shows database
- .env contains UPSTASH_REDIS_URL
- Test connection succeeds
</verify>
<done>Redis database created and configured</done>
</task>
<!-- NO CHECKPOINT NEEDED - Claude automated everything and verified programmatically -->
Example 3: Stripe Webhooks (Correct)
<!-- Claude automates everything -->
<task type="auto">
<name>Configure Stripe webhooks</name>
<files>.env, src/app/api/webhooks/route.ts</files>
<action>
1. Use Stripe API to create webhook endpoint pointing to /api/webhooks
2. Subscribe to events: payment_intent.succeeded, customer.subscription.updated
3. Save webhook signing secret to .env
4. Implement webhook handler in route.ts
</action>
<verify>
- Stripe API returns webhook endpoint ID
- .env contains STRIPE_WEBHOOK_SECRET
- curl webhook endpoint returns 200
</verify>
<done>Stripe webhooks configured and handler implemented</done>
</task>
<!-- Human verifies in Stripe dashboard -->
<task type="checkpoint:human-verify" gate="blocking">
<what-built>Stripe webhook configured via API</what-built>
<how-to-verify>
Visit Stripe Dashboard > Developers > Webhooks
Confirm: Endpoint shows https://myapp.com/api/webhooks with correct events
</how-to-verify>
<resume-signal>Type "yes" if correct</resume-signal>
</task>
Anti-Patterns
❌ BAD: Asking human to automate
<task type="checkpoint:human-action" gate="blocking">
<action>Deploy to Vercel</action>
<instructions>
1. Visit vercel.com/new
2. Import Git repository
3. Click Deploy
4. Copy deployment URL
</instructions>
<verification>Deployment exists</verification>
<resume-signal>Paste URL</resume-signal>
</task>
Why bad: Vercel has a CLI. Claude should run vercel --yes.
✅ GOOD: Claude automates, human verifies
<task type="auto">
<name>Deploy to Vercel</name>
<action>Run `vercel --yes`. Capture URL.</action>
<verify>vercel ls shows deployment, curl returns 200</verify>
</task>
<task type="checkpoint:human-verify">
<what-built>Deployed to {url}</what-built>
<how-to-verify>Visit {url}, check homepage loads</how-to-verify>
<resume-signal>Type "approved"</resume-signal>
</task>
❌ BAD: Too many checkpoints
<task type="auto">Create schema</task>
<task type="checkpoint:human-verify">Check schema</task>
<task type="auto">Create API route</task>
<task type="checkpoint:human-verify">Check API</task>
<task type="auto">Create UI form</task>
<task type="checkpoint:human-verify">Check form</task>
Why bad: Verification fatigue. Combine into one checkpoint at end.
✅ GOOD: Single verification checkpoint
<task type="auto">Create schema</task>
<task type="auto">Create API route</task>
<task type="auto">Create UI form</task>
<task type="checkpoint:human-verify">
<what-built>Complete auth flow (schema + API + UI)</what-built>
<how-to-verify>Test full flow: register, login, access protected page</how-to-verify>
<resume-signal>Type "approved"</resume-signal>
</task>
❌ BAD: Asking for automatable file operations
<task type="checkpoint:human-action">
<action>Create .env file</action>
<instructions>
1. Create .env in project root
2. Add: DATABASE_URL=...
3. Add: STRIPE_KEY=...
</instructions>
</task>
Why bad: Claude has Write tool. This should be type="auto".
Summary
Checkpoints formalize human-in-the-loop points. Use them when Claude cannot complete a task autonomously OR when human verification is required for correctness.
The golden rule: If Claude CAN automate it, Claude MUST automate it.
Checkpoint priority:
- checkpoint:human-verify (90% of checkpoints) - Claude automated everything, human confirms visual/functional correctness
- checkpoint:decision (9% of checkpoints) - Human makes architectural/technology choices
- checkpoint:human-action (1% of checkpoints) - Truly unavoidable manual steps with no API/CLI
See also: references/cli-automation.md for exhaustive list of what Claude can automate.