Skip to content

AI-Assisted Workflow

Complete workflow for contributing to Kartuli using AI assistance, from idea to merged PR.

Overview

The workflow has 4 phases:

  1. Refinement: Idea → Issue - Refine idea into GitHub issue
  2. Implementation: Issue → PR - Implement the issue
  3. Integration: PR → Main - Review and merge
  4. Cleanup - Post-merge housekeeping

Each phase can be done manually or with AI assistance.

Refinement: Idea → Issue

Manual

  1. Think through requirements
  2. Define acceptance criteria
  3. Create issue using GitHub template
  4. Check [x] label boxes
  5. Labels auto-apply via GitHub Action

With AI

  1. Use AI assistance to refine ideas and generate GitHub issues
    • Phase 1: Discuss idea with AI to refine requirements and acceptance criteria
    • Phase 2: Ask AI to format as GitHub issue with proper labels
  2. Create issue on GitHub
    • Review AI-generated markdown
    • Copy/paste into GitHub issue template
    • Labels auto-apply via GitHub Action

Result: Issue created in project board (Backlog status)

Implementation: Issue → PR

Create Branch

Option A: GitHub auto-create (recommended)

  • Click "Create a branch" in issue sidebar
  • Auto-linked to issue
  • Share issue URL with AI to find branch

Option B: Manual creation

  • Create branch following naming convention
  • Link PR to issue using Closes #X

Implement

Use AI assistance for implementation:

With Cursor or similar AI agent:

  1. Share issue URL with AI
  2. AI finds linked branch and switches to it
  3. AI guides through: docs → implement → test → commit → PR
  4. Confirm at decision points (commit message, PR updates)
  5. Copy PR content AI generates

Result: PR created, labels auto-propagate from issue

Handle Bot Feedback Loop

Automated code review bots (like Qodo) may provide suggestions. This can loop multiple times:

For each review cycle:

  1. Share PR link with AI agent after it's created
  2. Review suggestions - AI helps analyze each bot suggestion
  3. Discuss & decide - Determine what to implement/reject/defer
  4. Implement approved changes:
    • AI makes changes on same branch
    • Test and verify changes work
    • AI offers to generate formatted response comment for the bot
    • You confirm, AI provides markdown to copy/paste to PR
    • AI proposes commit message
    • You confirm, AI commits and pushes
  5. Wait for next bot review and repeat

Bot response format (AI generates this for you):

  • Lists each suggestion with team decision (✅ Applied / ❌ Rejected / ⏸️ Deferred)
  • Includes team's reasoning for each decision
  • Provides external AI perspective on the decisions made

Integration: PR → Main

  1. Review PR yourself
  2. Address any human feedback
  3. Merge when ready

Result: Issue + PR auto-close and move to "Done"

Cleanup

bash
git checkout main
git pull origin main
git branch -d {branch-name}

Project Board Tracking

StatusWhenHow
BacklogIssue createdAuto
ReadySpec completeManual
In ProgressPR openedAuto (maybe - test this)
DonePR mergedAuto

Quick Reference