AI-PHISR Documentation

A product management operating system built on Claude Code + Obsidian. Persistent context, specialized agents, 21 skills, 8 playbooks.

How It Works

Every interaction follows the same loop:

Prerequisites

Installation

Step 1: Install Node.js LTS

Install Node.js LTS from nodejs.org, then confirm both Node and npm are available in your terminal.

# Windows PowerShell
node --version
npm --version

# Mac/Linux
node --version
npm --version

Step 2: Install Claude Code

npm install -g @anthropic-ai/claude-code
claude --version

Step 3: Open the vault in Obsidian

Open Obsidian, click "Open folder as vault", and select the vault/ directory inside the ai-phisr project.

Step 4: Verify

# Windows PowerShell
cd "C:\path\to\ai-phisr"
Get-Location
dir
claude

# Mac/Linux
cd /path/to/ai-phisr
claude

Claude Code should start in the project root and accept input. The exact startup text may differ by version, so focus on behavior: Claude opens, slash commands are available, and you are in the same folder as CLAUDE.md.

Setup Worksheet

Before using the system, fill out Setup_Worksheet.md in the project root. It takes about 15 minutes and captures:

• Product overview — name, description, target users, key problems
• Strategy — vision, 1-3 strategic bets, OKRs
• Active initiatives — what you're building now
• Team & stakeholders — who you work with and what they care about
• Dependencies — cross-team coordination needs
• Communication style — how you write (for PM Voice Creator)
• Current sprint — what's happening right now

First Run

After filling out the worksheet, run the bootstrap command:

/populate-strategy

/populate-strategy is the setup/bootstrap command. It is separate from the 21 calibrated skills and is meant to create your initial strategy and vault structure.

This reads your worksheet and generates:

• Vision, strategic bets, and OKR files in vault/Strategy/
• Initiative files in vault/Initiatives/
• Stakeholder map in vault/Team/
• Context index in vault/System/memory/

5 AI Agents

Agents are specialized personas that route your request to the right skills. You don't need to invoke them directly — they activate based on what you ask.

21 Skills

Skills are slash commands you type in Claude Code. Each one loads your vault context, follows a specific workflow, and produces structured output.

8 Playbooks

Playbooks chain multiple skills into multi-step workflows for common PM scenarios. One command triggers the entire sequence.

Vault Structure

The Obsidian vault is organized by purpose. All files are plain Markdown with YAML frontmatter.

vault/
  Strategy/         Vision, strategic bets, OKRs
  Initiatives/      Active projects with status tracking
  Hypotheses/       Testable bets with validation status
  Decisions/        Decision log with context + rationale
  Research/
    personas/       User personas from research
    insights/       Synthesized research findings
  Team/             Stakeholders, dependencies, rituals
  Sprints/          Sprint docs, retros, incident reports
  System/
    memory/         Session log, sprint state, context index
    templates/      Note templates for Obsidian
  Archive/          Completed or deprecated items

Frontmatter Examples

Every vault file uses YAML frontmatter for structured data:

# Initiative frontmatter
---
title: "Smart Priority Engine"
status: active        # active | paused | completed | cancelled
strategic-bet: "AI-Powered Task Prioritization"
owner: "Sarah Kim"
priority: P0          # P0 | P1 | P2
target-date: 2026-03-31
---

# Hypothesis frontmatter
---
title: "AI Priority Trust"
initiative: "Smart Priority Engine"
status: testing       # untested | testing | validated | invalidated
risk-level: high      # high | medium | low
---

Memory Layer

Three files in vault/System/memory/ maintain persistent context across sessions:

Common Workflows

Morning Routine

Start your day by asking Claude about priorities. It reads your sprint state, active initiatives, and recent sessions automatically.

You: "What should I focus on today?"
Claude: [reads sprint state, surfaces blocked items, upcoming deadlines, open decisions]

Stakeholder Drops a Request

You: /playbook-new-idea
You: "The CEO wants us to add a Gantt chart view"
Claude: [runs 7-step workflow: frames problem, checks context, extracts assumptions, builds hypothesis]

Preparing for 1:1 with Manager

You: /playbook-manager-1on1
Claude: [scans initiatives, finds open decisions, builds decision-focused agenda in <2 min]

Best results come after your active initiatives, open decisions, manager info, and sprint state are in place.

End of Sprint

You: /sprint-summarizer
Claude: [summarizes progress, blockers, outcomes, observations]

You: /retro-facilitator
Claude: [facilitates structured retro with action items]

You: /playbook-sprint-kickoff
Claude: [starts next sprint with full context]

/playbook-sprint-kickoff works best once strategic bets, OKRs, dependencies, sprint state, and recent sprint history exist in the vault.

Need to Convince a Stakeholder

You: /playbook-pitch
You: "I need to convince the CEO to fund the AI prioritization bet for another quarter"
Claude: [audience analysis, narrative arc, data grounding, counter-arguments, slide outline]

5-Day Setup Guide (Half-Sprint)

Get AI-PHISR running on your real product in one work week.