Skip to content

udecode/dotai

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

██████╗  ██████╗ ████████╗ █████╗ ██╗
██╔══██╗██╔═══██╗╚══██╔══╝██╔══██╗██║
██║  ██║██║   ██║   ██║   ███████║██║
██║  ██║██║   ██║   ██║   ██╔══██║██║
██████╔╝╚██████╔╝   ██║   ██║  ██║██║
╚═════╝  ╚═════╝    ╚═╝   ╚═╝  ╚═╝╚═╝

Agent Instructions

Uses skiller to apply the same rules to all coding agents. Instructions and skills in .claude/rules/. npx skiller@latest apply generates CLAUDE.md/AGENTS.md, Skills and MCP configurations.

Quick Start

Fastest Setup

npx shadcn@latest add https://raw.githubusercontent.com/udecode/dotai/main/registry/all.json

Then, restart claude.

This installs all project files and configures Claude Code to auto-install the marketplace and plugins.

Manual Setup

Add marketplace and plugins (interactive):

/plugin marketplace add udecode/dotai

Restart Claude Code, then install all plugin files:

/dotai:install-all

Available Plugins

🛠️ dotai

Complete development toolkit - documentation, PRDs, debugging, PR workflows, and planning.

Installation:

/plugin install dotai@dotai
# restart claude
/dotai:install

Features:

  • Documentation - Generate and maintain app design and tech stack docs
  • PR Management - Full PR creation with reviews and draft mode
  • Planning - Implementation planning and log monitoring

Key Commands:

/dotai:create-app-design       # Generate design documentation
/dotai:opus <prompt>           # Run task with Opus model

Full Plugin Documentation →

🎓 skills

Meta-skills for finding, using, and writing Agent Skills - enforces skill usage protocols and provides skill authoring guidance.

Installation:

/plugin install skills@dotai
# restart claude

Features:

  • Using Skills - Mandatory protocols for skill discovery and usage
  • Writing Skills - TDD approach to skill authoring
  • No Rationalization - Prevents common excuses for skipping skills

Key Command:

/skills:skills   # Enforce skill usage protocols

Skills (auto-invoked):

  • using-skills - Mandatory workflows for finding and using skills

Full Plugin Documentation →

📋 plan

Planning and brainstorming workflows for software development - helps refine ideas into designs and create detailed implementation plans.

Installation:

/plugin install plan@dotai
# restart claude

Features:

  • Brainstorming - Transform rough ideas into fully-formed designs
  • Writing Plans - Create comprehensive implementation plans with TDD approach
  • Executing Plans - Batch execution with review checkpoints

Key Commands:

/plan:brainstorm      # Interactive design refinement (Opus)
/plan:write-plan     # Create implementation plan (Opus)
/plan:execute-plan   # Execute plan with checkpoints

Skills (auto-invoked):

  • brainstorming - Refine ideas through collaborative questioning
  • writing-plans - Create detailed task breakdowns
  • executing-plans - Execute plans in controlled batches

Full Plugin Documentation →

✅ prompt

Dynamic prompt injection system with before-start and before-complete checklists.

Installation:

The prompt system is automatically installed with the dotai registry (included in Quick Start setup). If you need to install it separately:

npx shadcn@latest add https://raw.githubusercontent.com/udecode/dotai/main/registry/prompt.json

This installs:

  • .claude/prompt.yml - Configuration file
  • .claude/scripts/user-prompt-submit.sh - Hook script for before-start/before-complete
  • .claude/scripts/post-compact.sh - Hook script for post-compact recovery
  • .claude/scripts/session-start.sh - Hook script for session start events

The hooks are automatically configured in .claude/settings.json when you install the dotai registry.

Features:

  • Before-Start Checklists - Enforce reminders before Claude responds
  • Before-Complete Checklists - Verification items before claiming completion
  • Post-Compact Recovery - Restore context after compaction with afterCompact instructions
  • Session Start - Load skills at session start (startup, resume, clear, compact)
  • Project-Specific - Configure different prompts per project

Configuration:

Edit .claude/prompt.yml:

beforeStart:
  - tag: MANDATORY-FIRST-RESPONSE
    header: 🚨 STOP - FIRST TOOL CALL MUST BE TodoWrite
    instructions:
      - DO NOT analyze. DO NOT read files. DO NOT edit.
      - "FIRST ACTION: TodoWrite with todo below as in_progress"
      - Condition NO → mark completed → proceed
    todos:
      - |
        Skill analysis (SKIP if 'quick' in message):
        (1) List ALL available skills
        (2) For EACH: YES/MIGHT/MAYBE = ✓
        (3) Skill(...) for ALL ✓ IN ONE PARALLEL CALL

beforeComplete:
  - tag: VERIFICATION-CHECKLIST
    header: 🔒 VERIFICATION REQUIRED
    instructions:
      - TodoWrite with ALL todos below
      - Condition NO → mark completed and skip
    todos:
      - "Typecheck (IF updated .ts files): Bash `bun typecheck`"
      - "Lint: Bash `bun lint:fix`"

afterCompact:
  - tag: POST-COMPACT-RECOVERY
    header: 🚨 CONTEXT WIPED - MANDATORY SKILL RELOAD
    instructions:
      - Context compaction DELETED all skills
      - DO NOT proceed until skill reload complete
    todos:
      - |
        Skill reload (MANDATORY):
        (1) List ALL available skills
        (2) Skill(...) for ALL applicable

Structure:

  • tag - Section identifier
  • header - Bold header shown at top of section
  • instructions - Operation guidelines (bulleted list)
  • todos - TodoWrite checklist items with conditional execution
  • afterCompact - Instructions after context compaction

🚀 agents

Agent orchestration patterns for parallel debugging and investigation - dispatch multiple Claude agents to solve independent problems concurrently.

Installation:

/plugin install agents@dotai
# restart claude

Features:

  • Parallel Dispatch - Multiple agents work concurrently on independent problems
  • Focused Scope - Each agent has narrow domain to investigate
  • Speed - Solve multiple problems in time of one

Key Command:

/agents:parallel   # Dispatch agents for parallel investigation

Skills (auto-invoked):

  • dispatching-parallel-agents - Orchestrate multiple agents for concurrent problem-solving

Full Plugin Documentation →

🔍 debug

Systematic debugging and root cause analysis framework - four-phase investigation process ensuring understanding before fixes.

Installation:

/plugin install debug@dotai
# restart claude

Features:

  • Systematic Debugging - Four-phase framework (investigate, analyze, test, implement)
  • Root Cause Tracing - Trace bugs backward through call stack
  • Defense-in-Depth - Add validation at multiple layers

Key Command:

/debug:debug   # Invoke systematic debugging framework

Skills (auto-invoked):

  • systematic-debugging - Four-phase debugging process
  • root-cause-tracing - Trace backward to find original trigger

Full Plugin Documentation →

🧪 test

Test-driven development workflow for writing tests before implementation - red-green-refactor cycle with deterministic unit tests.

Installation:

/plugin install test@dotai
# restart claude

Features:

  • Red-Green-Refactor - Classic TDD workflow
  • Smart Test Strategy - Auto-decide when to test based on complexity
  • Deterministic Tests - Only unit tests, no complex mocking

Key Command:

/test:tdd   # Test-driven development workflow

Skills (auto-invoked):

  • test-driven-development - Write test first, watch it fail, make it pass

Full Plugin Documentation →

🔀 git

Git and GitHub workflow automation - streamlined PR creation, draft management, and code review workflows.

Installation:

/plugin install git@dotai
# restart claude

Features:

  • Pull Request Creation - Full PR workflow with automatic incremental reviews
  • Draft PR Management - Efficient draft workflow without reviews
  • Smart Commits - Conventional commit messages and branch naming

Key Commands:

/git:create-pr   # Create PR with automatic code review
/git:draft-pr    # Create draft PR for work-in-progress

Skills (auto-invoked):

  • creating-pr - PR creation with automatic incremental reviews
  • drafting-pr - Draft PR management without reviews

Full Plugin Documentation →

🔔 notification

macOS notifications for Claude Code events.

Installation:

/plugin install notification@dotai
# restart claude

Features:

  • Get notified when Claude finishes responding
  • Ping on compact events

Full Plugin Documentation →

🎵 media

Auto-play/pause media on prompt submit/stop.

Installation:

brew install media-control  # Required
/plugin install media@dotai
# restart claude

Features:

  • Auto-play media when submitting prompts
  • Auto-pause media when stopping responses
  • Works with Spotify, Apple Music, YouTube, etc.

Full Plugin Documentation →

Workflows

1. Initial Setup

After Quick Start, create foundational docs:

# Create foundational docs
/dotai:create-app-design
/dotai:create-tech-stack

2. Daily Workflow

# Plan feature implementation with Opus
/plan:brainstorm      # Refine ideas into designs
/plan:write-plan      # Create implementation plan

/clear

# Start monitoring logs/errors
/watch

# Execute plan with checkpoints with Sonnet
/plan:execute-plan

# Work on features...
# Skills auto-load based on task context

# Create PR
/git:create-pr

Development

See CONTRIBUTING.md.

License

MIT

About

Context manager for all agents

Topics

Resources

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •