Skip to main content
Goal: Use AI to handle the cognitive overhead of documenting and delivering the work.

​
Automated Accounting

The AI handles the cognitive gear-shift of writing descriptive commit messages and Pull Request (PR) descriptions. Why this matters: Your brain is in “implementation mode”—switching to “documentation mode” is expensive.

​
Writing Commit Messages

Ask the AI to generate commit messages based on the chat history.

​
Example Prompt

Please write a conventional commit message for these changes.
Format: <type>(<scope>): <subject>

Types: feat, fix, refactor, docs, test, chore
Keep subject under 72 characters.
Include a body that explains the "why" behind the changes.

​
What Good Commit Messages Look Like

Bad: 
fix: updates
Good: 
feat(auth): add JWT refresh token rotation

Implements automatic token rotation to improve security.
Refresh tokens now expire after 7 days and are rotated
on each use. Adds middleware to handle token validation
and renewal transparently.

Closes #123

​
Automated Message Generation

For work done in a single chat session: 
"Based on our conversation, generate a commit message
that summarizes what we implemented and why."
For work spanning multiple sessions: 
"Run 'git diff main' and summarize all changes on this
branch into a comprehensive commit message."

​
Contextual Commits

For work spanning multiple sessions, ensure the AI has full context.

​
The Problem

AI only sees the current chat—it doesn’t know about earlier sessions.

​
The Solution

Prompt the AI to run git diff to see all changes: 
"Run 'git diff main...HEAD' to see all changes on this branch.
Then write a commit message that covers everything, not just
what we discussed in this session."

​
What This Prevents

  • Incomplete commit messages
  • Missing context from earlier work
  • Commits that don’t reflect full scope

​
Writing PR Descriptions

Generate comprehensive PR descriptions from the full context.

​
PR Description Template

"Please write a pull request description with:

## Summary
Brief overview of what this PR does

## Changes
Detailed list of changes made

## Testing
How this was tested

## Screenshots (if applicable)
Any visual changes

## Checklist
- [ ] Tests pass
- [ ] Documentation updated
- [ ] No breaking changes (or documented)

​
Example Prompt

"Based on our complete conversation and the git diff,
please write a comprehensive PR description following
the template above. Focus on the 'why' and 'what',
not just the 'how'."

​
Using GitHub CLI (gh)

Leverage the AI to interact with gh (GitHub CLI) for automated workflows.

​
Creating PRs Automatically

"Use the gh CLI to:
1. Create a new PR from this branch
2. Use the PR description we just wrote
3. Add labels: 'enhancement', 'needs-review'
4. Request review from @teammate
5. Link to issue #123"

​
What the AI Generates

gh pr create \
  --title "feat(auth): add JWT refresh token rotation" \
  --body "$(cat pr-description.md)" \
  --label enhancement,needs-review \
  --reviewer teammate \
  --web
Pro tip: Review the command before execution (remember: shell commands can’t be undone).

​
Agentic Mechanics

Use AI to automate the “boring” parts of the finish line:

​
Automate These Tasks

  • ✅ Write commit messages
  • ✅ Generate PR descriptions
  • ✅ Update changelogs
  • ✅ Create release notes
  • ✅ Update documentation
  • ✅ Run final checks (tests, linting)

​
Keep Human Control Over

  • ❌ Merging to main/master
  • ❌ Production deployments
  • ❌ Breaking changes
  • ❌ Security-sensitive operations

​
Pre-Ship Checklist

Before shipping, verify:
  • All tests pass locally
  • Linting passes
  • Type checking passes
  • No debug statements or console.logs
  • Git diff reviewed manually
  • Commit message is descriptive
  • PR description is comprehensive
  • Breaking changes are documented
  • CI/CD pipeline will pass
If anything fails: Fix before shipping, don’t “fix it later.”

​
The Final Push

​
Standard Workflow

# AI helps generate
git commit -m "feat(auth): add JWT refresh token rotation

Implements automatic token rotation for improved security.
[full description]"

# AI helps create PR
gh pr create --web

# Human reviews and merges
# (Keep final merge control)
Key Principle: AI handles documentation overhead. You maintain control over delivery decisions. Automate the boring, control the critical.