Metis

A structured system for AI-assisted product management. Skills, scripts, slash commands, and knowledge files that give any AI coding agent persistent context about your product and direct access to your tools.

Works with Claude Code, Cursor, Codex, Antigravity, or any agent that can read markdown files and run shell commands. The core is just structured markdown and Node.js scripts -- no vendor lock-in.

What's Inside

Category	Count	Examples
Core Skills	17	Daily ops, Jira, Confluence, SQL, coaching, weekly updates, Slack triage, interview prep
Expert Personas	13	Serial CEO, Principal PM, VC Investor, Growth Strategist, UX Psychologist, Devil's Advocate
Specialized Tools	10	PDF processing, video analysis, OCR, prototyping, transcript organization
Workflows	5	Expert panels, feature design debates, PR reviews, Jira-Confluence sync, research-to-doc
Customer Personas	4	Test features from realistic user perspectives
Utilities	6	Health checks, file management, system maintenance
Slash Commands	24	/pm-daily, /pm-coach, /pm-jira, /expert-panel, etc.
Integration Scripts	34	Google Workspace, Jira, Confluence, Slack, Dovetail, Granola, and more

Quick Start

# 1. Clone
git clone https://github.com/kyler-ross/mimir.git
cd mimir

# 2. Set up your AI instructions
cp CLAUDE.md.example CLAUDE.md
# Edit CLAUDE.md -- fill in the [bracketed] sections with your product details

# 3. Install script dependencies
cd scripts && npm install && cd ..

# 4. Add your credentials
cp scripts/.env.example scripts/.env
# Edit scripts/.env -- add your API keys

# 5. Add your product context
cp knowledge/product-overview.md.example knowledge/product-overview.md
cp knowledge/metrics-catalog.md.example knowledge/metrics-catalog.md
cp knowledge/team-members.json.example knowledge/team-members.json
# Edit each file with your real product info

# 6. Launch with your preferred tool
claude .                # Claude Code
cursor .                # Cursor
codex .                 # Codex
# Or point any AI agent at this directory

Connect Your Tools

Each integration is optional. Start with what you use and add more later.

Google Workspace (Calendar, Gmail, Sheets, Drive, Docs)

Requires a Google Cloud project with OAuth 2.0 credentials. One-time setup, then tokens auto-refresh.

node scripts/google-auth-setup.cjs

Jira + Confluence

# In scripts/.env
ATLASSIAN_EMAIL=you@company.com
JIRA_API_KEY=your-api-token
JIRA_BASE_URL=https://yourcompany.atlassian.net
CONFLUENCE_BASE_URL=https://yourcompany.atlassian.net/wiki

Get an API token at https://id.atlassian.com/manage-profile/security/api-tokens

Slack

# In scripts/.env
SLACK_BOT_TOKEN=xoxb-your-token

MCP Servers (PostHog, Figma, GitHub)

Copy and configure the MCP examples:

cp .claude/mcp.json.example .claude/mcp.json   # For Claude Code
cp .cursor/mcp.json.example .cursor/mcp.json   # For Cursor

See docs/mcp-setup.md for details on each server.

How It Works

Instruction files (CLAUDE.md, config/cursor-rules.md) set the rules -- tool priorities, safety constraints, output style, and references to skills and knowledge files. Most AI coding tools read these automatically from the repo root.

Skills (skills/) are structured instruction files following the SKILL.md standard. Each one tells the agent how to perform a specific task with real methodology. Any tool that can read markdown can use them.

Slash commands (.claude/commands/) are shortcuts that load skills. Tools that support custom commands (Claude Code, Cursor) pick these up automatically. For other tools, you can reference skills directly by path.

Scripts (scripts/) connect your agent to your tools via CLI. Calendar, Jira, Slack, Sheets -- all accessible from the terminal. Any agent that can run shell commands can use these.

Knowledge files (knowledge/) give your agent persistent context about your product, team, and metrics so you don't re-explain your business every conversation.

You: /pm-daily
        |
        v
Slash command loads --> Daily Chief of Staff skill
        |
        v
Skill instructs Claude to:
  1. Check calendar (scripts/google-calendar-api.js)
  2. Review Jira board (scripts/atlassian-api.cjs)
  3. Check Slack (scripts/slack-api.cjs)
  4. Cross-reference with product context (knowledge/)
        |
        v
Agent delivers: prioritized briefing with action items

Tool Compatibility

The repo includes config for multiple tools out of the box:

Tool	Instruction File	Commands	MCP Config
Claude Code	CLAUDE.md	.claude/commands/	.claude/mcp.json
Cursor	.cursorrules (symlink)	.cursor/commands/ (symlink)	.cursor/mcp.json
Codex / Others	Read CLAUDE.md or config/cursor-rules.md directly	Reference skills by path	N/A

The skills, scripts, and knowledge files are plain markdown and Node.js -- they work with anything that can read files and run commands.

Hooks (Claude Code)

The .claude/hooks/ directory includes pre-built hooks for Claude Code:

• CLI validation -- catches credential leaks, validates script arguments, detects wrong markup formats
• PR scope enforcement -- blocks multi-scope PRs and suggests how to split them
• Agent activation -- auto-suggests relevant skills based on your prompt
• MCP telemetry -- tracks tool usage for analytics
• Session startup -- health checks and personalized greeting

Copy .claude/settings.json.example to .claude/settings.json to enable them.

Customize It

Add Skills

mkdir -p skills/core/my-skill/references
# Write skills/core/my-skill/SKILL.md
# Write skills/core/my-skill/references/manifest.json
node scripts/generate-index.cjs

Add Slash Commands

Create .claude/commands/my-command.md:

Load and follow the skill at skills/core/my-skill/SKILL.md

User's request: $ARGUMENTS

Add Knowledge

Drop markdown files into knowledge/ and reference them in your CLAUDE.md.

Architecture

mimir/
├── CLAUDE.md.example          # AI instruction template (copy to CLAUDE.md)
├── config/cursor-rules.md     # Cursor-specific rules (symlinked from .cursorrules)
├── skills/                    # 55 skills across 6 categories
│   ├── _index.json            # Auto-generated routing index
│   ├── core/                  # Core PM skills (17)
│   ├── experts/               # Expert personas (13)
│   ├── specialized/           # Specialized tools (10)
│   ├── workflows/             # Multi-skill workflows (5)
│   ├── personas/              # Customer personas (4)
│   └── utilities/             # System utilities (6)
├── .claude/
│   ├── commands/              # 24 slash commands
│   ├── hooks/                 # CLI validation, PR enforcement, agent activation
│   ├── agents/                # Agent routing rules
│   ├── mcp.json.example       # MCP server config template
│   └── settings.json.example  # Hook configuration template
├── .cursor/
│   ├── commands -> ../.claude/commands
│   └── mcp.json.example
├── scripts/                   # 34 integration scripts + 15 client libraries
│   ├── lib/                   # Shared client libraries
│   ├── .env.example           # Credential template
│   └── package.json
├── knowledge/                 # Product context templates
├── docs/                      # Setup guides
└── examples/                  # Reference material

Requirements

• Any AI coding agent that can read files and run shell commands
• Node.js 18+
• Python 3.9+ (for OCR and transcript tools)
• API keys for the integrations you want to use

License

MIT