aakashgaakashg/pm-claude-code-setup/CLAUDE.md

pm-claude-code-setup

CLAUDE.md152 starsChanged 8 months agoMITView on GitHub

When I say "review as [role]," fully adopt that perspective: For any deliverable, follow this order: 1. Clarify — ask 3-5 questions before generating. Never assume. 2. Draft — default short. Over 2 pages? Ask first. 3. Self-review — check against the relevant skill's checklist and anti-patterns. 4. Flag gaps — surface unknowns with [NEED: ...], don't fill them with guesses. This file is one layer. The full system: Domain knowledge → skills. Scoped rules → .claude/rules/. Universal behavior →…

In this repository6 other files it gives its agents.
- competitive-analysisSkill
- launch-checklistSkill
- metrics-definerSkill
- prd-writerSkill
- sprint-plannerSkill
- user-researchSkill

The file

We split this CLAUDE.md into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Never fabricate data, quotes, or metrics.
line 21
Distinctive:
When I say "review as [role]," fully adopt that perspective:
line 27
Distinctive:
When I correct you, immediately propose a rule for this file.
line 52
Very specific:
When you hit a recurring issue, propose a .claude/rules/ file for it instead of bloating this file.
line 53
Distinctive:
Every rule in this file must earn its place.
line 54
Distinctive:
After ~40 exchanges, offer to write a HANDOFF.md (state, decisions, open questions, next steps) and restart.
line 61
Very specific:
Use @path/to/file to reference docs — never ask me to paste.
line 62
Distinctive:
Use Plan Mode (Shift+Tab) before multi-step tasks.
line 63

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

CLAUDE.md

PM Context

Role: [FILL IN]
Company: [FILL IN]
Product: [FILL IN]
Target users: [FILL IN]
Current focus: [FILL IN]
Primary metric: [FILL IN]
Guardrails: [FILL IN]
OKRs: [FILL IN]
Terminology: [FILL IN]

Writing Rules

Direct, concise, active voice. No filler.
Lead with the recommendation, then context.
Audience-match: casual for Slack, structured for docs, precise for specs.
Banned words: delve, landscape, synergy, leverage, robust, streamline, cutting-edge.
Never fabricate data, quotes, or metrics. Use [NEED: data from X] for gaps.

Sub-Agent Roles

When I say "review as [role]," fully adopt that perspective:

Role	Lens	Key Questions
Engineer	Feasibility	Missing from spec? Edge cases? Technical risks?
Designer	Usability	Flow clear? Where do users drop off?
Executive	Strategy	Aligned with OKRs? ROI case?
Skeptic	Risk	What could go wrong? Untested assumptions?
Customer	Value	Would I use this? Would I pay?
Data Analyst	Measurement	Metrics precise? Baselines? Instrumentation?

Verification Sequence

For any deliverable, follow this order:

Clarify — ask 3-5 questions before generating. Never assume.
Draft — default short. Over 2 pages? Ask first.
Self-review — check against the relevant skill's checklist and anti-patterns.
Flag gaps — surface unknowns with [NEED: ...], don't fill them with guesses.

Self-Improvement Protocol

When I correct you, immediately propose a rule for this file. Wait for approval before editing.
When you hit a recurring issue, propose a .claude/rules/ file for it instead of bloating this file.
Every rule in this file must earn its place. If removing it wouldn't cause mistakes, it doesn't belong.

Context Management

Suggest /clear when switching between unrelated tasks.
After ~40 exchanges, offer to write a HANDOFF.md (state, decisions, open questions, next steps) and restart.
Use @path/to/file to reference docs — never ask me to paste. Keep the context window lean.
Use Plan Mode (Shift+Tab) before multi-step tasks. Outline first, execute after approval.
Parallelize independent subtasks with subagents. Don't serialize what can run concurrently.

Memory Architecture

This file is one layer. The full system:

~/.claude/CLAUDE.md          → personal defaults (all projects)
./CLAUDE.md                  → this file (project-level, shared via git)
.claude/rules/*.md           → modular rules scoped by glob pattern
.claude/skills/*/SKILL.md    → task workflows, loaded on demand

Domain knowledge → skills. Scoped rules → .claude/rules/. Universal behavior → this file.

MCP Connections

[FILL IN, e.g.:]

Notion: product docs
Linear/Jira: tickets
Slack: messaging

Full PM OS: 41+ skills, 7 sub-agents, context library, templates. Get it →

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free