Coding standards and quality gates for product managers working with AI assistants in shared codebases.
PMs using AI to code face a unique challenge: shipping value quickly while not creating cleanup work for engineering teams. AI-assisted coding can be incredibly productive, but without guardrails, it often results in:
- PRs that fail CI/CD
- Code that doesn't match team patterns
- Tech debt that slows down the team
- Frustrated engineers cleaning up after you
A set of practical guidelines for PMs to code responsibly with AI assistance, whether working solo or in shared codebases with senior engineers.
pm-who-codes.md - Core philosophy and principles
- Role clarity: PM who codes vs. software engineer
- Working in shared codebases respectfully
- Task breakdown and documentation-first approach
- When to ask vs. when to ship
quality-gates.md - Pre-commit checklist
- What to check before every commit
- Checkpoint strategy for frequent commits
- How to initialize coding sessions with AI
- CI/CD readiness standards
solo-project-standards.md - Standards for solo projects
- Keep it simple, lean, and maintainable
- The 6-month test (will future you understand this?)
- When to test, when to document
- Pre-ship checklist
session-management.md - Managing context and continuity
- When to restart sessions (avoiding context rot)
- How to document your work before restarting
- Where to save session notes (team vs. solo)
- Starting fresh sessions without losing context
- PMs learning to code with AI assistance
- PMs working in shared codebases with engineering teams
- Product builders who want to ship responsibly with maintainable code
- Anyone using AI to code who wants to avoid sloppy code that fails CI or doesn't match patterns
Add to your context files:
These files work with any Claude Code setup. Choose what works for you:
A) Global context - Available in all projects:
# Add to your global Claude directory
cp *.md ~/.claude/B) Project-specific context - For a specific codebase:
# Add to your project's .claude directory
cp *.md /path/to/your-project/.claude/C) Reference in instructions:
In your CLAUDE.md or project instructions, reference these guidelines:
## Coding Standards
When working on code:
- Follow pm-who-codes.md for philosophy and approach
- Use quality-gates.md checklist before every commit
- Apply solo-project-standards.md for solo projects
- Use session-management.md to maintain continuity across sessionsKeep these files open while coding and review as needed:
- Before starting: Read pm-who-codes.md
- Before committing: Check quality-gates.md
- For solo projects: Follow solo-project-standards.md
- After 2-3 tasks: Follow session-management.md to restart
Fork this repo and adapt the guidelines to your team's:
- Specific tools and workflows
- Testing requirements
- Code review processes
- Architectural patterns
Just ask Claude to guide you:
I'm working on [feature] for [project].
Review the pm-coding-guardrails files and guide me through this coding session.That's it. Claude will:
- Read the relevant guardrail files
- Tell you what to focus on first
- Guide you through checkpoints
- Remind you about quality gates before commits
- Tell you when to restart sessions
If you want more control over what Claude focuses on, use these prompts:
Starting a team project:
I'm working on [feature] in a shared codebase with senior engineers.
Context:
- Review pm-who-codes.md (team project section)
- Review quality-gates.md for commit standards
- Match existing patterns exactly
Help me find similar code to match, then break this into small tasks.Before committing:
Before I commit, let's verify using quality-gates.md:
- Pattern matching
- Formatters/linters
- Tests written and passing
- All CI checks passed locally
Walk me through the checklist.Restarting after context rot:
Starting fresh session.
Context:
- Review session-management.md for approach
- Read .claude/sessions/[date]-[feature].md from last session
- Review pm-who-codes.md and quality-gates.md
Confirm you understand where we left off, then continue.Solo project:
Working on my solo project [name].
Context:
- Review solo-project-standards.md
- Review quality-gates.md
Guide me through this session.- Consistency > Personal Preference - Match existing patterns exactly
- CI Must Pass Locally First - Never push code that fails checks
- Quality > Speed - Shipping sloppy code isn't shipping value
- Test Before Ship - Untested code is broken code
- Know When to Ask - Better to ask than create cleanup work
- Discover Before Building - Search for existing patterns before adding new code
These guidelines incorporate best practices from experienced engineers working with AI:
- Tiny task breakdown - Keep AI laser-focused on small, discrete tasks
- Frequent checkpoints - Commit after every small task passes
- Context window management - Restart sessions after 2-3 big tasks
- Documentation first - Write markdown docs before coding
- Many-shot examples - Initialize sessions with patterns to follow
This isn't about being a perfect engineer. It's about being a responsible one.
The goal: Ship features that create value without creating work for others.
I need to add user authentication to the dashboard.
Context:
- Match existing patterns in components/auth/
- Follow quality-gates.md checklist
- Break into small tasks per pm-who-codes.md
- Commit after each task passes checks
Let's start by reading similar components and breaking this down.I'm a PM working in a codebase with senior engineers.
Important:
- Study similar code first, match patterns exactly (pm-who-codes.md)
- Run all CI checks locally before committing (quality-gates.md)
- Ask before deviating from established conventions
- Commit frequently after small, working changes
Let's ensure I don't create cleanup work for the team.Found these helpful? Have improvements? PRs welcome.
These guidelines synthesize practices from experienced product managers and engineers:
- Core philosophy developed through hands-on PM coding work
- Best practices incorporate feedback from senior engineers working with AI
- Checkpoint and session management strategies adapted from production engineering teams
CC BY-NC-SA 4.0 - Free to use for personal and internal business use with attribution.
Built for PMs who want to code responsibly with AI. Ship features with quality and intention.
