god-mode

name	god-mode
description	God's eye view of your dev repos. Multi-project tracking across GitHub/Azure DevOps. AI learns from your commits to upgrade your agents.md.
metadata	{"openclaw":{"requires":{"bins":["gh","sqlite3","jq"]}}}
user-invocable	true

god-mode Skill

Conversational project oversight and AI agent coaching for OpenClaw.

Overview

god-mode lets users ask about their projects in plain English. You run the commands and explain the results conversationally.

What users can ask:

• "Set up god-mode for my repos"
• "What's happening across my projects?"
• "Which repos need attention?"
• "Analyze my agents.md"
• "What should I work on today?"

What you provide:

• Multi-project status summaries (GitHub + Azure DevOps)
• Plain-English explanations of commit activity, PRs, issues
• LLM-powered agent instruction analysis
• Recommendations for improving their AGENTS.md

Perfect for:

• Developers juggling multiple repos who want conversational oversight
• Teams using AI coding assistants (Claude, Copilot, etc.)
• Anyone who wants data-driven improvements to their AGENTS.md

Getting Started

For OpenClaw Users (Conversational)

Just tell me what you want! I'll run the god-mode commands for you.

Example conversations:

You: "Set up god-mode for my tada repository"

Me: Sure! I'll set up god-mode and add your tada repo. [runs god setup and god projects add github:YourUsername/tada] Done! Your repo is configured. Want me to sync the data?

You: "Yes, and show me what's happening"

Me: [runs god sync and god status]

Here's your overview:

• tada: 30 commits this week, last activity 3 days ago
• 0 open PRs, 0 issues

Your most active project! Commits are all features and bug fixes.

You: "Analyze my agents.md for tada"

Me: [runs god agents analyze github:YourUsername/tada]

I've analyzed your AGENTS.md against 155 commits. Here's what I found:

GAPS:

• Testing practices (high impact) - 68 bug fixes but no testing guidance
• Voice API debugging - 12 commits mention "voice" but no troubleshooting

STRENGTHS:

• TypeScript strictness is working - zero type-related fixes

Want me to add testing guidance to your AGENTS.md?

What you can ask me:

• "Set up god-mode for [repo]"
• "Show me all my projects"
• "What's happening across my repos?"
• "Analyze my agents.md"
• "Which repos need attention?"
• "What should I work on today?"

I'll handle all the commands and explain the results in plain English.

For CLI Users (Direct Commands)

If you prefer running commands yourself:

# Setup
god setup

# Add a project
god projects add github:username/repo

# Sync and view
god sync
god status

Commands

god status [project]

Show overview of all projects, or details for one:

god status              # All projects
god status myproject    # One project in detail

god sync [project] [--force]

Fetch/update data from repositories:

god sync                # Incremental sync all
god sync myproject      # Just one project
god sync --force        # Full refresh (ignore cache)

god projects

Manage configured projects:

god projects                        # List all
god projects add github:user/repo   # Add project
god projects remove myproject       # Remove project

god review [--month YYYY-MM]

Generate monthly activity reviews:

god review                  # Last month's activity
god review --month 2026-01  # Specific month
god review --json           # JSON output

What it shows:

• Total commits across all projects
• Most active repositories
• Pull request activity (merged, active, closed)
• Detailed breakdown by project with date ranges
• Perfect for monthly retrospectives and planning

Example output:

Monthly Review: 2026-01
  📊 286 commits across 7 projects
  👥 10 unique contributors
  
Most Active Projects:
  tada - 155 commits
  ContentEngine - 63 commits
  brain - 27 commits

Use cases:

• Monthly team stand-ups
• Personal retrospectives ("What did I actually work on?")
• Quarterly planning ("Which projects got attention?")
• Automated monthly summaries via cron

god agents analyze <project>

Analyze agents.md against commit history using LLM:

god agents analyze myproject

What it does:

1. Fetches your AGENTS.md from the repository
2. Analyzes commit patterns (types, pain points, frequently changed files)
3. Calls an LLM (Claude/GPT) to find gaps and suggest improvements
4. Displays recommendations interactively
5. Optionally applies changes to your AGENTS.md

LLM Configuration:

god-mode automatically detects and uses the best available LLM:

1. OpenClaw (default when running as skill) - Uses your OpenClaw agent
2. Anthropic - Set ANTHROPIC_API_KEY="sk-ant-..."
3. OpenAI - Set OPENAI_API_KEY="sk-..."
4. OpenRouter - Set OPENROUTER_API_KEY="sk-or-..."
5. Manual - Outputs prompt if no LLM available

When running in OpenClaw:

• The analysis prompt is displayed to your OpenClaw agent
• You (or your agent) provides the JSON analysis directly in the conversation
• Much simpler than managing separate API keys!

OpenClaw Workflow:

When you run god agents analyze in OpenClaw:

1. Analysis starts:

   🔭 Analyzing github:InfantLab/tada
   ✅ Found AGENTS.md (remote)
   ✅ 155 commits analyzed
   🤖 Using OpenClaw's LLM
   

2. I (OpenClaw agent) receive the analysis prompt showing:

   • Your complete AGENTS.md content
   • Commit pattern summary (45 features, 68 bug fixes, etc.)
   • Most changed files/directories
   • Pain points and commit samples

3. I analyze and provide JSON response:

   {
     "gaps": [
       {
         "area": "Testing",
         "observation": "68 bug fixes but no testing guidance in AGENTS.md",
         "impact": "high",
         "suggestion": "Add testing section with coverage targets"
       }
     ],
     "strengths": [...],
     "recommendations": [...]
   }
   

4. god-mode displays results and offers to apply changes to your AGENTS.md

5. You choose which recommendations to accept, and god-mode updates the file

Standalone Workflow (outside OpenClaw):

If you set ANTHROPIC_API_KEY or OPENAI_API_KEY, god-mode calls the API directly:

export ANTHROPIC_API_KEY="sk-ant-..."
god agents analyze myproject  # Fully automated

god agents generate <project> (Coming Soon)

Bootstrap agents.md for a new project by analyzing repo structure.

god logs [options]

View activity logs:

god logs                # Last 50 lines
god logs -n 100         # Last 100 lines
god logs -f             # Follow log output
god logs --path         # Show log file location
god logs --clear        # Clear all logs

All god-mode activity is logged to ~/.god-mode/logs/activity.log with timestamps for transparency and debugging.

Configuration

Config file: ~/.config/god-mode/config.yaml

projects:
  - id: github:user/repo
    name: My Project      # Display name
    priority: high        # high/medium/low
    tags: [work, api]
    local: ~/code/myrepo  # Local clone path

sync:
  initialDays: 90         # First sync lookback
  commitsCacheMinutes: 60

analysis:
  agentFiles:             # Files to search for
    - agents.md
    - AGENTS.md
    - CLAUDE.md
    - .github/copilot-instructions.md

Data Storage

All data stored locally in ~/.god-mode/:

• cache.db - SQLite database (commits, PRs, issues, analyses)
• contexts/ - Saved workspace contexts (v0.2)

Authentication

god-mode uses your existing CLI authentication:

Provider	CLI	Setup
GitHub	gh	gh auth login
Azure	az	az login
GitLab	glab	glab auth login

No tokens stored by god-mode. We delegate to CLIs you already trust.

Requirements

• gh - GitHub CLI (for GitHub repos)
• sqlite3 - Database
• jq - JSON processing

Examples

Morning Check-In

god status
# See all projects at a glance
# Notice any stale PRs or quiet projects

Before Switching Projects

god status myproject
# See recent activity, open PRs, issues
# Remember where you left off

Improving Your AI Assistant

god agents analyze myproject
# Get suggestions based on your actual commit patterns
# Apply recommendations to your agents.md

Weekly Review

god status
# Review activity across all projects
# Identify projects needing attention

Agent Guide (For OpenClaw Agents)

When the user asks about their projects, here's what to do:

"Set up god-mode" / "Track my [repo]"

1. Run god setup (checks dependencies)
2. Run god projects add github:username/repo
3. Run god sync to fetch data
4. Summarize what you found

"Show me my projects" / "What's happening?"

1. Run god status
2. Translate output to conversational summary:
   • "Your most active repo is X with Y commits this week"
   • "Z repo hasn't had activity in N days"
   • "You have M open PRs across all projects"

"Analyze my agents.md"

1. Run god agents analyze github:username/repo
2. You'll receive the analysis prompt with:
   • Their complete AGENTS.md
   • Commit pattern analysis (155 commits, 68 bug fixes, etc.)
   • Pain points and frequently changed files
3. Provide JSON analysis in this format:

   {
     "gaps": [
       {"area": "Testing", "observation": "68 bug fixes but no test guidance", "impact": "high", "suggestion": "Add testing section"}
     ],
     "strengths": [
       {"area": "TypeScript", "observation": "Zero type errors in 155 commits"}
     ],
     "recommendations": [
       {"priority": 1, "section": "## Testing", "content": "- Write unit tests for new code\n- Run tests before commits"}
     ],
     "summary": "Strong TypeScript practices, needs testing guidance"
   }
   

4. Summarize the analysis conversationally
5. Offer to apply recommendations

"Which repos need attention?"

1. Run god status
2. Look for:
   • Stale PRs (>3 days old)
   • No activity in >5 days
   • Open issues piling up
3. Suggest what to focus on

Automated Workflows

Daily Briefing (Heartbeat):

# HEARTBEAT.md
- Run `god status` and summarize:
  - Projects with stale PRs (>3 days)
  - Projects with no activity (>5 days)
  - Open PRs needing review

Monthly Review (Cron - 1st of month):

schedule:
  kind: cron
  expr: "0 9 1 * *"  # 9am on 1st of each month
  tz: "America/New_York"
payload:
  kind: agentTurn
  message: |
    Run god review for last month and summarize:
    - Which projects were most active?
    - Any projects that went quiet?
    - Major accomplishments from commit messages
    - Recommendations for next month
sessionTarget: isolated

Weekly Analysis (Cron):

schedule: "0 9 * * 1"  # Monday 9am
task: |
  Run `god agents analyze` on high-priority projects.
  If gaps found, notify with suggestions.

Common Questions

How do I use god-mode?

In OpenClaw: Just ask me! "Show me my projects", "Analyze my agents.md", etc. I'll run the commands and explain the results.

CLI: Run commands directly: god status, god sync, god agents analyze

Do I need to set up anything first?

First time: Tell me to "set up god-mode for [your repo]" and I'll handle it. Or run god setup and god projects add github:your/repo yourself.

Authentication: Make sure gh auth login is done (GitHub CLI authentication).

Do I need an API key?

No! When you ask me to analyze your agents.md, I receive the analysis prompt and provide the JSON response directly. No separate API key needed.

Standalone: If using god-mode outside OpenClaw, you can set ANTHROPIC_API_KEY or OPENAI_API_KEY for automated analysis.

How often does it sync data?

When you ask: I run god sync when you ask about your projects. First sync fetches 90 days of history. Subsequent syncs are incremental (only new data).

Manual: You can run god sync anytime, or god sync --force for a full refresh.

What data gets stored?

Locally only: Commits, PRs, issues, and analysis results in ~/.god-mode/cache.db. Activity logs in ~/.god-mode/logs/activity.log. Nothing sent to external servers (except when calling LLM APIs if configured).

Does it work with private repos?

Yes! Uses your gh CLI authentication, so it has access to whatever your GitHub account can access.

Troubleshooting

"gh: command not found"

Install GitHub CLI: https://cli.github.com/

"Not logged in to GitHub"

Run: gh auth login

"No projects configured"

Add a project: god projects add github:user/repo

Stale data

Force refresh: god sync --force

Agent analysis returns empty {}

This is normal in OpenClaw mode - the prompt is displayed for the OpenClaw agent to analyze. The agent provides the JSON response in conversation, not as return value.

---

OpenClaw Community Skill
License: MIT
Repository: https://github.com/InfantLab/god-mode-skill