01
Feature Clarification
01-feature-definition.md
Feed raw Slack messages, meeting notes, or a one-liner. Claude rewrites it as a clean feature definition with user stories and success criteria.
↓ raw request
↑ 01-feature-definition.md
EXAMPLE INPUT →
"Clients want to get reminded before their pet appointment" →
Claude produces: Feature Definition + 3 user stories + 5 success criteria
Claude produces: Feature Definition + 3 user stories + 5 success criteria
Paste the feature definition .md into Linear or Jira. AI auto-generates ticket title, description, acceptance criteria, and story points estimate.
↓ 01-feature-definition.md
↑ Ticket + story points
EXAMPLE →
Linear AI reads the user stories → creates sub-tasks: "Backend: reminder scheduler", "Frontend: notification settings screen", "DB: reminder_preferences table"
Auto-create a feature wiki page from the .md. Notion AI formats it with headings, summary block, and links it to the project space.
↓ 01-feature-definition.md
↑ Wiki page (auto-structured)
EXAMPLE →
Paste .md → Notion AI creates a page with: Overview, Problem Statement, Success Criteria, and links to related pages automatically
↓
01-feature-definition.md feeds into →
02
Check Existing Flow
02-flow-map.md
The .md contains a Mermaid flowchart block. Paste into mermaid.live or Eraser.io — instantly renders an interactive user flow diagram. No manual drawing.
↓ 02-flow-map.md (Mermaid block)
↑ Visual flow diagram
EXAMPLE →
Claude wrote: flowchart TD → BookingScreen → ReminderSettings → SchedulerService → NotificationQueue → PushNotification. Paste into mermaid.live → shareable diagram link in 10 seconds.
Paste the flow map into FigJam. FigJam AI converts bullet-point flows into sticky-note wireframe maps. Design team can immediately annotate and refine.
↓ 02-flow-map.md
↑ FigJam sticky flow board
EXAMPLE →
Entry points list → FigJam AI creates a board with screens as nodes, arrows as flows, and highlights shared components in a different color
Feed the dependency map into Cursor. Ask it to find all files in the codebase that match the listed dependencies and flag which ones will be modified.
↓ 02-flow-map.md (dependency table)
↑ File list + affected lines
EXAMPLE →
"Dependencies: BookingService, NotificationQueue" → Cursor searches codebase → highlights booking.service.ts line 142 and queue.worker.ts line 67 as the exact files to modify
↓
02-flow-map.md feeds into →
03
Impact Analysis
03-risk-mitigation-plan.md
Feed the impact list to Claude with your DB schema. Claude identifies N+1 query risks, missing indexes, and cascade delete dangers before a single line is written.
↓ 03-impact-list.md + DB schema
↑ Annotated risk register
EXAMPLE →
"Impact: appointments table (RW)" → Claude flags: "appointments has 2M rows, adding reminder_sent column needs zero-downtime migration with DEFAULT NULL first, then backfill"
Convert the risk register table into a living Notion database. Coda AI auto-assigns risk owners from your team roster and creates a follow-up task for each High risk.
↓ 03-risk-mitigation-plan.md
↑ Risk tracker DB with owners
EXAMPLE →
Risk table → Notion DB with: Risk, Severity, Owner (auto-assigned by role), Status (Open/Mitigated), Due Date
The impact list names every affected DB table. Feed these into dbdiagram.io — Claude generates the DBML schema snippet. Visualize the table relationships before design starts.
↓ 03-impact-list.md (DB section)
↑ ER diagram (pre-design)
EXAMPLE →
"appointments table + new reminder_schedules table" → Claude generates DBML → paste into dbdiagram.io → visual ER diagram ready for design review
↓
03-impact-list.md feeds into →
04
Feature Design
04-frontend-design.md
Paste the frontend-design.md screen descriptions into v0. It generates React + Tailwind UI components for each screen. Copy directly into your codebase.
↓ 04-frontend-design.md
↑ React UI components
EXAMPLE →
"Screen: Reminder Settings. Components: toggle list (SMS/Email/Push), time picker, frequency selector, save button" → v0 generates a complete styled React settings screen
Feed screen names + component list + user flow into Galileo AI. It generates high-fidelity UI mockups from text. Designer refines instead of building from scratch.
↓ 04-frontend-design.md
↑ High-fidelity UI mockups
EXAMPLE →
Paste screen description → Galileo generates: mobile reminder settings screen with toggle switches, time picker, and confirmation modal — all pixel-perfect, ready to hand off
Feed the database-design.md schema into Claude with your ORM. Claude generates the full Prisma schema or Drizzle table definitions — ready to copy into your project.
↓ 04-database-design.md
↑ schema.prisma / drizzle tables
EXAMPLE →
DB design describes reminder_schedules table → Claude generates Prisma model with relations to appointments and users, including @index decorators for the query patterns described
Convert the backend-design.md API flow into a Mermaid sequence diagram. Shows exactly which service calls which, in what order — useful for async flows and queues.
↓ 04-backend-design.md
↑ Sequence diagram
EXAMPLE →
API endpoints + flow → Claude generates: sequenceDiagram Client → API → ReminderService → Queue → Worker → PushService → Notification sent. Paste into mermaid.live instantly.
↓
04-design-doc.md feeds into →
05
Development
05-pr-summary.md
Open Cursor, paste the full design doc. Use Composer mode to generate all files at once — controllers, services, routes, components — following your existing code style.
↓ 04-design-doc.md
↑ All feature files generated
EXAMPLE →
Cursor Composer: "Generate all files for this feature" → creates reminder.controller.ts, reminder.service.ts, reminder.routes.ts, ReminderSettings.tsx, reminder.test.ts in one shot
Use the design doc as the system prompt for Claude Code. It writes the implementation, runs it, fixes errors, and iterates — all in the terminal without context switching.
↓ 04-backend-design.md
↑ Runnable code + self-fixed bugs
EXAMPLE →
Claude Code reads design doc → writes POST /reminders endpoint → runs tests → notices missing auth middleware → adds it → all without leaving the terminal
The pr-summary.md becomes the GitHub PR description automatically. Copilot for PRs reads the diff + summary and writes structured change notes with impact callouts.
↓ 05-pr-summary.md
↑ PR description + review checklist
EXAMPLE →
PR summary .md → GitHub PR auto-fills: "What changed, Why, DB migrations included, Tests added, How to test locally" — reviewer gets full context in one read
↓
05-pr-summary.md + test cases feed into →
06
Testing & Regression
06-test-cases.md
Feed test-cases.md into Claude. It writes Playwright end-to-end test scripts for every scenario in the table — happy path, edge cases, and failure states.
↓ 06-test-cases.md
↑ playwright.spec.ts files
EXAMPLE →
TC-001 "User enables SMS reminders, saves settings" → Claude generates: test('should save SMS reminder preference', async () => { ... }) — complete Playwright script, runnable immediately
Feed the security checklist .md into Snyk or ZAP. They scan the new endpoints against each checklist item and produce a pass/fail security report automatically.
↓ 06-security-audit.md
↑ Security scan report
EXAMPLE →
Checklist: "Endpoints require auth" → Snyk tests POST /reminders without token → confirms 401 returned → adds to report as PASS
Performance test thresholds from the .md become Datadog monitor configs. AI generates the monitor YAML from the threshold table — auto-alerting from day one.
↓ 06-test-report.md (perf section)
↑ Monitor config + alert rules
EXAMPLE →
"p95 target: 200ms for GET /reminders" → Claude generates Datadog monitor YAML: alert if p95 > 300ms for 5 min → ready to apply to infrastructure
↓
06-test-report.md feeds into →
07
Release & Deployment
07-release-notes.md
The deployment-checklist.md becomes a GitHub Actions workflow. Claude converts each checklist item into a CI step — smoke tests, DB migration check, health ping — all automated.
↓ 07-deployment-checklist.md
↑ deploy.yml workflow file
EXAMPLE →
Checklist: "Run smoke tests → Check migration ran → Ping health endpoint → Enable feature flag" → Claude generates the complete GitHub Actions deploy.yml with all steps
Release notes .md → Slack AI formats it as a #releases announcement. Intercom AI converts it into a customer-facing in-app announcement with the right tone for pet business owners.
↓ 07-release-notes.md
↑ Slack post + in-app announcement
EXAMPLE →
Technical release notes → Intercom AI rewrites as: "🐾 New: Appointment reminders! Your clients now get SMS and email reminders automatically. No setup needed." — customer-ready in seconds
Monitoring dashboard .md defines metrics + thresholds. Claude generates the Datadog dashboard JSON and PagerDuty alert policy — live monitoring from the moment of deploy.
↓ 07-monitoring-dashboard.md
↑ Dashboard JSON + alert policy
EXAMPLE →
Monitoring table → Claude generates Datadog dashboard with: reminder queue depth, delivery success rate, API error rate — all pre-wired to PagerDuty alerts