PRISM - AI-Powered PM Capability Toolkit Rules

Project: PRISM (Lăng kính) - formerly Root v2.1 Version: 1.0 Platform: Claude Code only Updated: 2025-11-28

🎯 Project Overview

PRISM is an AI-powered PM capability toolkit that provides on-demand product management competencies through Claude Code slash commands.

Concept: A prism to help Product Managers see the full spectrum of possibilities within a problem.

Core Philosophy:

Capability over complexity: Frameworks are tools, not rules
Synthesis over signposting: Provide answers, not file paths
Keywords over links: Enable self-research, don't hand-hold
Context continuity: Track conversations to maintain coherence
Independent capabilities: Each command works standalone without prerequisites
User freedom: No workflow enforcement

📁 Project Structure

prism/
├── PRISM-CONVOS/              # Conversation tracking (mandatory)
│   ├── 00-CONVO-LOGS.md       # Conversation index
│   └── *.md                   # Individual conversations
├── data/                      # PM work products
│   ├── items/                 # Categorized PM items
│   ├── prds/                  # Product Requirements Documents
│   ├── roadmaps/              # Roadmap files
│   ├── metrics/               # Analytics dashboards
│   ├── retrospectives/        # Retrospective records
│   └── scores/                # Prioritization scores
├── .claude/
│   ├── commands/              # Slash commands (10 total)
│   │   ├── assist.md
│   │   ├── scenario/*.md      # 7 scenario commands
│   │   └── gh/*.md            # 2 git helpers
│   └── product-foundation/    # 45 PM frameworks (HIDDEN)
│       ├── COMPETENCY-MAP.md
│       ├── SCENARIO-INDEX.md
│       └── [framework files]
├── CLAUDE.md                  # Claude Code guidance
├── RULES.md                   # This file
├── README.md                  # User documentation
├── REVAMPING.md               # Migration specification
└── .gitignore                 # Git ignore rules

🤖 Claude Code Rules

1. Conversation Management (MANDATORY)

BEFORE ANY RESPONSE, Claude MUST:

Check for existing conversations:
- Read PRISM-CONVOS/00-CONVO-LOGS.md
- Search by topic keywords from user input
- Look for matching conversations in the index
If matching conversation found:
- Read the full conversation file
- Continue the existing conversation (append new messages)
- Update date_modified in YAML frontmatter
- DO NOT create duplicate conversations
- Maintain context from previous exchanges
If no matching conversation found:
- Generate filename: yyyy-mm-dd-topic-keywords.md
- Create file in PRISM-CONVOS/ with YAML frontmatter
- Add entry to 00-CONVO-LOGS.md immediately
- Update search index

Conversation file format:

---
conversation_id: "yyyy-mm-dd-topic-name"
date_created: "yyyy-mm-dd"
date_modified: "yyyy-mm-dd"
topic: "Topic Description"
status: "active"
participants:
  - role: "user"
    name: "User"
  - role: "assistant"
    platform: "claude-code"
tags:
  - "tag1"
  - "tag2"
summary: "Brief summary"
related_files: []
---

# Conversation content starts here

**User** (yyyy-mm-dd HH:MM):
[User message]

**Claude** (yyyy-mm-dd HH:MM):
[Assistant response]

Conversation status values:
- active: Ongoing conversation
- completed: Resolved/finished
- archived: Historical reference

2. Framework Synthesis Protocol (CRITICAL)

Frameworks in .claude/product-foundation/ are Claude's reference library, NOT user-facing documentation.

How Claude Must Use Frameworks:

✅ Read frameworks silently
- Users never see Claude reading files
- Frameworks are hidden in .claude/product-foundation/
- No file paths shown to users
✅ Synthesize content
- Extract core methodology (steps, formulas, models)
- Pull out real examples from frameworks
- Identify common mistakes/pitfalls
- Adapt to user's specific context
✅ Provide direct guidance
- Give actionable steps
- Explain "how to do it", not "read about it"
- Include specific examples
- Offer concrete advice
✅ Share keywords for research
- Provide 5-7 keywords/concepts
- Enable independent learning
- Point to research areas, not files
❌ Never say "read this file"
- Users can't access .claude/product-foundation/
- Don't recommend browsing frameworks
- Don't show file paths
❌ Never show framework file paths
- Keep .claude/product-foundation/ invisible
- Frameworks are Claude's knowledge base
- Users get insights, not references

Synthesis Response Template:

## 🎯 I Understand Your Scenario
[Restate user's goal clearly]

## 📚 Framework Approach: [Framework Name]

**Core methodology**:
[Explain the framework's approach - steps, formulas, models]

**How to apply it**:
[Specific guidance for their scenario]

**Key concepts**:
- [Principle 1]
- [Principle 2]
- [Principle 3]

**Common pitfalls**:
- [Mistake 1]
- [Mistake 2]

## ⚡ Quick Action Plan

- [ ] [Specific action from framework]
- [ ] [Follow-up action]
- [ ] [Validation step]

**Time estimate**: [Duration]

## 🎓 Key Takeaway

[One sentence summarizing the most important insight]

## 💡 Pro Tips

[2-3 practical tips from framework]

## 🔍 Keywords for Further Research

- [Keyword 1]
- [Keyword 2]
- [Keyword 3]
- [Keyword 4]
- [Keyword 5]

3. Project References

Project name: Always "PRISM" (never "Root" or "Productify")
Version: PRISM v1.0
Framework directory: .claude/product-foundation/ (hidden)
Conversation directory: PRISM-CONVOS/
Data directory: data/
Platform: Claude Code only (no multi-platform references)

4. Anti-Hallucination Protocol

Never invent data or assume completion
Validate directory structures before claiming tasks done
Use evidence-based progress reporting
Cross-reference with existing files
Always prompt for missing arguments (never hallucinate values)
Read files before editing them
Verify paths exist before operations

📋 Framework Usage Rules

Framework Categories (45 Total)

Located in .claude/product-foundation/:

Discovery & Research (6 frameworks):

Jobs-to-Be-Done (JTBD)
Persona Development
Value Proposition Canvas
Opportunity Solution Tree
Double Diamond
Design Thinking & Lean Startup

Prioritization (3 frameworks):

RICE Prioritization
ICE Scoring
MoSCoW Prioritization

Validation (2 frameworks):

Product-Market Fit (PMF)
Kano Model

Analysis (4 frameworks):

Market Sizing (TAM/SAM/SOM)
Cost-Benefit Analysis & ROI
MECE Framework
Balanced Tradeoff Matrix

Strategy (3 frameworks):

Product Roadmapping
SWOT Analysis
Product-Led Growth (PLG)

Documentation (2 frameworks):

PRD Template
User Story Mapping

Execution (1 framework):

OKRs (Objectives & Key Results)

Launch (1 framework):

Go-to-Market (GTM) Strategy

Optimization (1 framework):

A/B Testing & Experimentation

Metrics (6 frameworks):

Product Metrics
North Star Framework
AARRR (Pirate Metrics)
Retention & Churn Analysis
Cohort Analysis
Feature Adoption

Analytics (1 framework):

Data Visualization & Charts

Psychology (1 framework):

Gamification Tactics

People Management (3 frameworks):

Interviewing Framework
Feedback Framework
One-on-One Framework

Communication (2 frameworks):

Stakeholder Management
Influence Without Authority

Domain-Specific:

FinTech (3): Financial Inclusion, Payment Systems, Business Models
TravelTech (2): Industry Overview, Optimization Strategies
UX/UI (3): UX Research, UX Design, UI Optimization

Framework Selection Logic

Claude should use COMPETENCY-MAP.md and SCENARIO-INDEX.md to route user scenarios:

Example mappings:

"prioritize features" → RICE Prioritization
"launch product" → Go-to-Market Strategy
"validate PMF" → Product-Market Fit
"reduce churn" → Retention & Churn Analysis
"interview candidate" → Interviewing Framework
"influence stakeholders" → Stakeholder Management or Influence Without Authority
"rural banking access" → FinTech Financial Inclusion
"booking conversion" → TravelTech Optimization

Framework Reading Requirements

When using a framework, Claude MUST:

Read the entire framework file (not just snippets)
Understand the full context and methodology
Extract relevant examples and case studies
Identify applicable common mistakes
Synthesize into actionable guidance
Provide keywords for deeper research

💬 Slash Command Rules

Available Commands (11 Total)

Smart Assistant:

/assist [description] - AI framework navigator

Scenario Commands:

/scenario:feedback - Giving/receiving feedback
/scenario:prioritize - Feature prioritization
/scenario:chart - Data visualization
/scenario:interview - PM interviewing
/scenario:launch - Product launch
/scenario:churn - Churn analysis
/scenario:pmf - Product-Market Fit

Git Helpers:

/gh:commit - Git commit helper
/gh:init - Git initialization

Help & Guidance:

/prism-help - First-time user guide with real-world examples

Command Behavior Requirements

ALL slash commands MUST:

Check conversation tracking first
- Read PRISM-CONVOS/00-CONVO-LOGS.md
- Search for existing conversations
- Continue or create as appropriate
Read frameworks silently
- Access .claude/product-foundation/ files
- Users never see this step
- Load complete framework content
Synthesize content
- Extract methodology
- Adapt to user context
- Provide direct guidance
Provide keywords
- 5-7 research keywords
- Enable independent learning
- No file paths
Give actionable steps
- Specific, concrete actions
- Extracted from frameworks
- Tailored to scenario

Command Argument Handling

Accept arguments when provided
Prompt for arguments when missing
Never hallucinate or improvise missing values
Validate inputs before processing
Show clear error messages for invalid inputs

📊 Data Storage Rules

Data Directory Structure

data/
├── items/              # Categorized PM items
│   ├── ideas/          # IDEA-NNN.md
│   ├── incidents/      # INC-NNN.md
│   ├── initiatives/    # INIT-NNN.md
│   ├── ops-requests/   # OPS-NNN.md
│   ├── csat/           # CSAT-NNN.md
│   ├── integrations/   # INTEG-NNN.md
│   ├── funnel/         # FUNNEL-NNN.md
│   ├── tech-debt/      # DEBT-NNN.md
│   ├── competitive/    # COMP-NNN.md
│   ├── compliance/     # COMPLY-NNN.md
│   ├── insights/       # INSIGHT-NNN.md
│   ├── research/       # RES-NNN.md
│   └── platform/       # PLAT-NNN.md
├── prds/              # PRD-NNN.md
├── roadmaps/          # roadmap-yyyy-qN.md
├── metrics/           # metrics-yyyy-mm.md
├── retrospectives/    # retro-yyyy-mm-dd.md
└── scores/            # score-feature-name.md

File Format Standards

All data files MUST use YAML frontmatter:

---
id: "PREFIX-NNN"
type: "category"
status: "new|active|completed|archived"
date_created: "yyyy-mm-dd"
date_modified: "yyyy-mm-dd"
priority: "low|medium|high|critical"
tags: []
related_items: []
assigned_to: ""
---

# Content starts here

Data File Naming

Items: PREFIX-NNN.md (e.g., IDEA-001.md, INC-042.md)
PRDs: PRD-NNN.md or PRD-product-name.md
Roadmaps: roadmap-yyyy-qN.md (e.g., roadmap-2025-q2.md)
Metrics: metrics-yyyy-mm.md
Retrospectives: retro-yyyy-mm-dd.md
Scores: score-feature-name.md

🔒 Quality Standards

Code Quality

Validate all file paths before operations
Use evidence-based claims (never assume completion)
Cross-reference documentation for consistency
Maintain bilingual content where applicable (English/Vietnamese)
Read files before editing
Never create files unnecessarily

Documentation Quality

Keep CLAUDE.md and README.md synchronized
Update conversation logs with all interactions
Maintain accurate progress tracking
Document all framework additions
Use consistent terminology (PRISM, not Root)
Update version markers appropriately

Framework Synthesis Quality

DO:

✅ Read frameworks completely before responding
✅ Extract actionable steps, not just theory
✅ Provide real examples from frameworks
✅ Share common mistakes from frameworks
✅ Give 5-7 research keywords
✅ Tailor guidance to user's specific scenario
✅ Maintain bilingual support

DON'T:

❌ Say "read this file"
❌ Show file paths to users
❌ Recommend browsing .claude/product-foundation/
❌ Reference frameworks by filename
❌ Assume user can access hidden directories
❌ Provide generic advice without framework synthesis

Conversation Quality

Check for existing conversations before creating new ones
Maintain context across conversation continuations
Update date_modified when continuing conversations
Use clear, specific topic names for new conversations
Add meaningful tags for searchability
Update summaries when conversations evolve

🚫 Anti-Patterns to Avoid

Framework Usage Anti-Patterns

❌ Bad: "Read .claude/product-foundation/PRIORITIZATION-rice-prioritization.md for feature prioritization"

✅ Good: "For feature prioritization, use the RICE framework:

Reach: How many users affected? (estimate 0-1000+)
Impact: Value per user (Minimal=0.25, Low=0.5, Medium=1, High=2, Massive=3)
Confidence: How certain are you? (Low=50%, Medium=80%, High=100%)
Effort: Person-months required

Score = (Reach × Impact × Confidence) / Effort

Prioritize features with highest RICE scores.

Keywords for research: RICE scoring, opportunity cost, value vs effort matrix, prioritization frameworks"

Conversation Management Anti-Patterns

❌ Bad: Creating duplicate conversations without checking logs

✅ Good: Always checking PRISM-CONVOS/00-CONVO-LOGS.md first

❌ Bad: Using generic topic names like "conversation-001.md"

✅ Good: Using descriptive names like "2025-11-28-feature-prioritization-saas.md"

Path Reference Anti-Patterns

❌ Bad: "The framework is in 03-PRODUCT-FOUNDATION/"

✅ Good: "Let me synthesize the framework guidance for you..."

❌ Bad: "You can find this in the old Root directory"

✅ Good: "Based on the [Framework Name] methodology..."

🎓 Best Practices

For Framework Synthesis

Always read the full framework before responding
Extract real examples from the framework content
Identify user's experience level and adapt language
Provide time estimates for actions
Include common mistakes from framework
Give specific steps, not vague advice
End with keywords for deeper research

For Conversation Management

Search thoroughly in conversation logs before creating new files
Use consistent topic naming: yyyy-mm-dd-topic-keywords.md
Update date_modified every time you continue a conversation
Add relevant tags for future searchability
Maintain YAML frontmatter integrity
Update summaries as conversations evolve

For User Interaction

Restate their scenario to confirm understanding
Be specific about which framework you're using (name it)
Provide actionable steps they can take immediately
Estimate time/effort for recommendations
Share keywords for independent learning
Ask clarifying questions when scenario is ambiguous

🔄 Version History

PRISM v1.0 (Current)

Date: 2025-11-28
Changes:
- Single platform (Claude Code only)
- Hidden frameworks (.claude/product-foundation/)
- Synthesis-focused approach (Claude reads, users get insights)
- Keyword-based research (users Google independently)
- Streamlined structure (removed multi-platform overhead)
- Production-ready (no development artifacts)
- Renamed from Root to PRISM

Root v2.1 (Previous)

Date: 2025-11-24
Status: Multi-platform (5 AI platforms)
Frameworks: 45 comprehensive PM frameworks
Commands: 65 command files across platforms

Root v2.0

Date: 2025-11-20
Change: Shifted from workflow management to capability toolkit

Productify (Original)

Date: 2024-early 2025
Approach: Workflow-based PM assistant

📝 File Naming Conventions

File Type	Pattern	Example
Slash commands	`category/command.md`	`.claude/commands/scenario/feedback.md`
Frameworks	`PREFIX-name.md`	`DISCOVERY-jobs-to-be-done.md`
Conversations	`yyyy-mm-dd-topic-keywords.md`	`2025-11-28-feature-prioritization.md`
Data items	`PREFIX-NNN.md`	`IDEA-001.md`, `INC-042.md`
PRDs	`PRD-NNN.md` or `PRD-name.md`	`PRD-001.md`, `PRD-mobile-app.md`
Roadmaps	`roadmap-yyyy-qN.md`	`roadmap-2025-q2.md`
Metrics	`metrics-yyyy-mm.md`	`metrics-2025-11.md`
Retrospectives	`retro-yyyy-mm-dd.md`	`retro-2025-11-28.md`
Scores	`score-feature-name.md`	`score-dark-mode.md`

🎯 Success Metrics

PRISM succeeds when:

✅ Users get unstuck and move forward with confidence
✅ Frameworks are synthesized into actionable guidance
✅ Users receive keywords for independent research
✅ Conversations maintain context across sessions
✅ No file paths are shown to users
✅ Guidance is specific and tailored to user scenarios
✅ Common mistakes are proactively addressed

Remember: PRISM is about capability, not complexity. Frameworks are Claude's reference library - users get synthesized insights and research keywords, never file paths. Always maintain conversation continuity through PRISM-CONVOS/.

FilesExpand file tree

RULES.md

Latest commit

History