██████╗ ██████╗ ████████╗ █████╗ ██╗
██╔══██╗██╔═══██╗╚══██╔══╝██╔══██╗██║
██║ ██║██║ ██║ ██║ ███████║██║
██║ ██║██║ ██║ ██║ ██╔══██║██║
██████╔╝╚██████╔╝ ██║ ██║ ██║██║
╚═════╝ ╚═════╝ ╚═╝ ╚═╝ ╚═╝╚═╝
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.
npx shadcn@latest add https://raw.githubusercontent.com/udecode/dotai/main/registry/all.jsonThen, restart claude.
This installs all project files and configures Claude Code to auto-install the marketplace and plugins.
Add marketplace and plugins (interactive):
/plugin marketplace add udecode/dotaiRestart Claude Code, then install all plugin files:
/dotai:install-allComplete development toolkit - documentation, PRDs, debugging, PR workflows, and planning.
Installation:
/plugin install dotai@dotai
# restart claude
/dotai:installFeatures:
- 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 modelMeta-skills for finding, using, and writing Agent Skills - enforces skill usage protocols and provides skill authoring guidance.
Installation:
/plugin install skills@dotai
# restart claudeFeatures:
- 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 protocolsSkills (auto-invoked):
using-skills- Mandatory workflows for finding and using skills
Planning and brainstorming workflows for software development - helps refine ideas into designs and create detailed implementation plans.
Installation:
/plugin install plan@dotai
# restart claudeFeatures:
- 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 checkpointsSkills (auto-invoked):
brainstorming- Refine ideas through collaborative questioningwriting-plans- Create detailed task breakdownsexecuting-plans- Execute plans in controlled batches
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.jsonThis 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 applicableStructure:
tag- Section identifierheader- Bold header shown at top of sectioninstructions- Operation guidelines (bulleted list)todos- TodoWrite checklist items with conditional executionafterCompact- Instructions after context compaction
Agent orchestration patterns for parallel debugging and investigation - dispatch multiple Claude agents to solve independent problems concurrently.
Installation:
/plugin install agents@dotai
# restart claudeFeatures:
- 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 investigationSkills (auto-invoked):
dispatching-parallel-agents- Orchestrate multiple agents for concurrent problem-solving
Systematic debugging and root cause analysis framework - four-phase investigation process ensuring understanding before fixes.
Installation:
/plugin install debug@dotai
# restart claudeFeatures:
- 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 frameworkSkills (auto-invoked):
systematic-debugging- Four-phase debugging processroot-cause-tracing- Trace backward to find original trigger
Test-driven development workflow for writing tests before implementation - red-green-refactor cycle with deterministic unit tests.
Installation:
/plugin install test@dotai
# restart claudeFeatures:
- 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 workflowSkills (auto-invoked):
test-driven-development- Write test first, watch it fail, make it pass
Git and GitHub workflow automation - streamlined PR creation, draft management, and code review workflows.
Installation:
/plugin install git@dotai
# restart claudeFeatures:
- 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-progressSkills (auto-invoked):
creating-pr- PR creation with automatic incremental reviewsdrafting-pr- Draft PR management without reviews
macOS notifications for Claude Code events.
Installation:
/plugin install notification@dotai
# restart claudeFeatures:
- Get notified when Claude finishes responding
- Ping on compact events
Auto-play/pause media on prompt submit/stop.
Installation:
brew install media-control # Required
/plugin install media@dotai
# restart claudeFeatures:
- Auto-play media when submitting prompts
- Auto-pause media when stopping responses
- Works with Spotify, Apple Music, YouTube, etc.
After Quick Start, create foundational docs:
# Create foundational docs
/dotai:create-app-design
/dotai:create-tech-stack# 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-prSee CONTRIBUTING.md.
MIT