Agentic Coding Workflow — Visualised

Step 1 of 5 — Feature Request
Step 1 — Feature request
Developer asks the agent to build a feature
The agent reads the full scope before doing anything
👤
Build a new FastAPI REST endpoint for ...
🤖
Got it. Before I write any code, let me break this into reviewable chunks and propose a plan for your approval.

Reading full scope...

Identifying layers and dependencies...
Step 2 — Chunk proposal
Agent proposes the breakdown — waits for approval
No code written yet
🤖
Here's my proposed chunk breakdown. Please review and confirm before I start:
#BranchContains
✓   Human approved the plan — writing FEATURE_PLAN.md
Step 3 — Execution
Agent executes one chunk at a time
One branch → one concern → one PR → confirm → next chunk
$ cat FEATURE_PLAN.md # re-read plan before each chunk
chunk 1: feat/db-schema ✓ confirmed
$ git town hack feat/db-schema
✓ Created feat/db-schema off main
... implementing schema and migrations only ...
$ git town propose
✓ PR opened: feat/db-schema → main
$ git town append feat/validators
✓ Created feat/validators stacked on feat/db-schema
... implementing validators only ...
$ git town propose
✓ PR opened: feat/validators → feat/db-schema
→ Repeating for service, controller...
Step 4 — Review
Stack is ready — small focused PRs for review
Each PR is one concern. Reviewable in minutes.
BranchCurrent branchTarget branchStatus
feat/controller feat/controller feat/service Open
feat/service feat/service feat/validators Open
feat/validators feat/validators feat/db-schema Open
feat/db-schema feat/db-schema main Open
Step 5 — Merge cascade
Merging bottom up — GitHub retargets automatically
No manual retargeting needed
BranchCurrent branchTarget branchStatus
feat/controller feat/controller feat/service Open
feat/service feat/service feat/validators Open
feat/validators feat/validators feat/db-schema Open
feat/db-schema feat/db-schema main Open