Beginner~15 minAI Workflows

AI Task Workflow

The everyday Knowns loop — create a task, plan, implement, and verify a feature using /kn-* skills.

TasksSkillsEveryday
/kn-init/kn-research/kn-plan/kn-implement
kn-init
kn-research
Create task
kn-plan
kn-implement
Done ✓

The everyday loop

This is the workflow you'll use most often. No spec needed — just a clear task and your AI assistant. Always research first so your AI understands the codebase before planning.

What You'll Build

  • A task with clear acceptance criteria
  • An AI-generated implementation plan
  • Working code implemented by your AI assistant
  • Tracked progress with time and notes

Initialize

/kn-init

AI loads your project context — existing tasks, docs, conventions, and memory.

Research the Area

Context before code

Always research before planning. Your AI writes better plans when it understands existing patterns, related code, and past decisions.

/kn-research health check endpoint, Express route pattern, monitoring
AI researching
AI:  Searching project context for "health check"...

   Found:
   - Code: src/routes/ — existing route pattern
   - Memory: "Express + TypeScript project"
   - No existing health endpoint

   Ready to create task with full context.

Create a Task

Ask your AI to create a task, or do it from the CLI:

AI creates the task
You: Create a task to add a health check endpoint
   that returns server status and uptime

AI:  ✓ Created task-7x2k: "Add health check endpoint"
   AC-1: GET /health returns 200 with status "ok"
   AC-2: Response includes server uptime in seconds
   AC-3: Response includes current timestamp

Or via CLI:

knowns task create "Add health check endpoint" \
  -d "Simple health check for monitoring" \
  --ac "GET /health returns 200 with status ok" \
  --ac "Response includes server uptime" \
  --ac "Response includes timestamp"

Plan

/kn-plan 7x2k

AI reads the task, searches related docs and code, and creates a plan:

Generated plan
AI:  Plan for task-7x2k:
   1. Create src/routes/health.ts
   2. Add GET /health route returning JSON
   3. Include process.uptime() and Date.now()
   4. Register route in app.ts
   5. Verify all 3 ACs pass

The plan is saved to the task so any future session can pick it up.

Implement

/kn-implement 7x2k
AI implementing
AI:  Starting task-7x2k: "Add health check endpoint"
   → Reading plan...
   → Memory: "Express + TypeScript project"

   Creating src/routes/health.ts...
   Updating src/app.ts to register route...

   ✓ AC-1: GET /health returns 200 ✓
   ✓ AC-2: Uptime included ✓
   ✓ AC-3: Timestamp included ✓
   ✓ All acceptance criteria met

Review the Result

knowns task 7x2k --plain
Task status
Task: Add health check endpoint
Status: done
ACs: [✓] GET /health returns 200
   [✓] Response includes uptime
   [✓] Response includes timestamp
Notes: Created health.ts, registered in app.ts

Tips

Be specific with ACs

"Returns 200 with JSON body" is better than "Works correctly". Clear ACs help your AI know exactly when it's done.

Let AI plan first

The plan step catches issues before coding starts. It's faster to fix a plan than to rewrite code.

Use time tracking

Run knowns time start 7x2k before you begin. Time data helps with sprint planning and retrospectives.

What's next?

Ready for a more structured workflow? Try the Spec-Driven Fullstack Auth example, or read about Task Management for the full lifecycle.