Skip to main content
Pro Workflow works with Cursor through the .cursor plugin system and .mdc rules. Skills, agents, and references are translated to Cursor’s format.

Plugin Manifest

Location: .cursor-plugin/plugin.json
{
  "name": "pro-workflow",
  "version": "2.0.0",
  "description": "Complete AI coding workflow system. Orchestration patterns, 18 hook events, 5 agents, 7 reference guides, cross-agent support, and searchable learnings.",
  "author": {
    "name": "Rohit Ghumare",
    "url": "https://github.com/rohitg00"
  },
  "homepage": "https://github.com/rohitg00/pro-workflow",
  "repository": "https://github.com/rohitg00/pro-workflow",
  "license": "MIT",
  "keywords": [
    "claude-code",
    "workflow",
    "productivity",
    "self-correction",
    "worktrees",
    "hooks",
    "agents",
    "best-practices",
    "sqlite",
    "fts5",
    "searchable"
  ],
  "skills": ["./skills/"],
  "agents": [
    "./agents/planner.md",
    "./agents/reviewer.md",
    "./agents/scout.md",
    "./agents/orchestrator.md",
    "./agents/debugger.md"
  ]
}

Installation

# Install SkillKit (cross-agent skill translator)
npm install -g skillkit

# Translate Pro Workflow to Cursor format
cd your-project
skillkit translate pro-workflow --agent cursor

Cursor Rules (.mdc)

Cursor uses .mdc (Markdown Context) files for rules. Pro Workflow provides rules in rules/.
# Pro Workflow Rules

## Self-Correction

When corrected or when a mistake is made:
1. Acknowledge specifically what went wrong
2. Propose a rule: `[LEARN] Category: Rule`
3. Wait for approval before adding to LEARNED section

## Planning

Use multi-step planning when:
- Task touches >5 files
- Architecture decisions needed
- Requirements unclear

## Quality Gates

Before marking complete:
1. Run lint
2. Run typecheck
3. Run tests (at least --related)
4. Check for console.log, TODOs, secrets

## Code Review

Pause for review at:
- Plan completion
- Every 5 file edits
- Before git operations
- Auth/security code

## Learnings

Capture learnings in this format:

Skills Translation

SkillKit translates Claude Code skills to Cursor-compatible format:
Source (Claude Code):
---
name: wrap-up
description: End-of-session ritual
user-invocable: true
allowed-tools: ["Bash", "Read"]
---

# Wrap-Up

Run these checks:
1. git status
2. Run tests
3. Capture learnings
Translated (Cursor):
# Wrap-Up Skill

End-of-session ritual.

## Usage

Run these checks:
1. git status
2. Run tests
3. Capture learnings

## Available in Cursor

Use this skill by asking: "Run wrap-up ritual"

Agents in Cursor

Cursor doesn’t have native agent support. Pro Workflow provides agent instructions as .mdc rules:

Planner Mode

Add to .cursor/planner.md:
# Planner Mode

When asked to plan:
1. Explore code (read-only)
2. List files to change
3. Present plan for approval
4. Never edit without approval

Reviewer Mode

Add to .cursor/reviewer.md:
# Reviewer Mode

When asked to review:
1. Check logic, edge cases, errors
2. Security audit
3. Performance analysis
4. Present findings

Database Integration

The Pro Workflow database works with Cursor via Node.js scripts:
// save-learning.ts
import { createStore } from 'pro-workflow';

const store = createStore();

store.addLearning({
  project: process.env.PROJECT_NAME || 'cursor-project',
  category: 'Testing',
  rule: 'Run tests before committing',
  mistake: 'Pushed broken code',
  correction: 'Added pre-commit hook'
});

store.close();
console.log('Learning saved');

Worktree Support

Cursor doesn’t have native worktree support. Use git worktrees manually:
# Create worktree for feature
git worktree add ../project-feature feature-branch

# Open in new Cursor window
cursor ../project-feature

# When done, remove worktree
git worktree remove ../project-feature

Hook Alternatives

Cursor doesn’t support hooks like Claude Code. Use git hooks instead:
# .git/hooks/pre-commit
#!/bin/bash

echo "Running quality gates..."

# Lint
if ! npm run lint; then
  echo "Lint failed. Fix errors before committing."
  exit 1
fi

# Type check
if ! npm run typecheck; then
  echo "Type check failed. Fix errors before committing."
  exit 1
fi

# Tests
if ! npm test -- --changed; then
  echo "Tests failed. Fix tests before committing."
  exit 1
fi

echo "Quality gates passed."
# .git/hooks/post-commit
#!/bin/bash

# Track commit in database
node -e "
const { createStore } = require('pro-workflow');
const store = createStore();
const sessionId = 'cursor-' + Date.now();
store.updateSessionCounts(sessionId, 1, 0, 0);
store.close();
"

echo "Commit tracked."

Configuration Files

{
  "cursor.proWorkflow.enabled": true,
  "cursor.proWorkflow.dbPath": "~/.pro-workflow/data.db",
  "cursor.proWorkflow.autoWrapUp": false,
  "cursor.proWorkflow.qualityGates": ["lint", "typecheck", "test"]
}

Cross-Agent Workflows

Use Claude Code for complex tasks, Cursor for quick edits:
TaskToolWhy
PlanningClaude CodeNative agent support
Quick fixesCursorFast tab completions
Multi-file refactorClaude CodeBetter orchestration
Single-file editsCursorInline editing
Code reviewClaude CodeReviewer agent
ExplorationBothParallel sessions
Use the same .mcp.json and ~/.pro-workflow/data.db across both tools. Learnings are shared!

Feature Comparison

FeatureClaude CodeCursor
Native Agents❌ (use rules)
Skills❌ (use rules)
Hooks❌ (use git hooks)
Commands/command⚠️ (custom scripts)
Worktrees✅ Native⚠️ Manual
MCP Servers
Database API
Tab Completion
Inline Edit
Multi-Phase Dev✅ Native⚠️ Manual

Next Steps

SkillKit

Translate skills to any agent

Claude Code Plugin

Full-featured Claude Code setup

Database API

Learnings and session tracking

Cross-Agent Guide

Use both tools together

Build docs developers (and LLMs) love