pro-workflow

Complete AI coding workflow system.
Orchestration patterns • 18 hook events • 5 agents • 7 reference guides • Cross-agent support
Works with Claude Code, Cursor, and 32+ agents via SkillKit.

What's New in v2.0

Orchestration Patterns	Command > Agent > Skill architecture with multi-phase development
5 Agents	planner, reviewer, scout, orchestrator (RPI workflow), debugger
18 Hook Events	Added SubagentStart/Stop, TaskCompleted, PermissionRequest, TeammateIdle, PostToolUseFailure
7 Reference Guides	Settings, CLI cheatsheet, orchestration patterns, context loading, cross-agent workflows, new features, daily habits
Context Optimizer	Token management and context budget planning skill
Production Settings	Full settings.example.json with permissions, spinner, output style
Curated MCP Config	Battle-tested server recommendations with scope guidance
/develop Command	Research > Plan > Implement > Review & Commit with validation gates
/doctor Command	Health check for your pro-workflow setup

How It Works

"80% of my code is written by AI, 20% is spent reviewing and correcting it." — Karpathy

Pro Workflow optimizes for that ratio. Every pattern reduces correction cycles.

The Self-Correction Loop

Corrections compound. Each mistake becomes a rule that prevents future mistakes. After 50 sessions, Claude barely needs correcting.

The /develop Flow

Multi-phase development with validation gates. Research before planning, plan before implementing, review before committing.

The /develop Command in Action

Architecture

Patterns

Pattern	What It Does
Self-Correction Loop	Claude learns from your corrections automatically
Parallel Worktrees	Zero dead time - native claude -w worktrees
Wrap-Up Ritual	End sessions with intention, capture learnings
Split Memory	Modular CLAUDE.md for complex projects
80/20 Review	Batch reviews at checkpoints
Model Selection	Opus 4.6 adaptive thinking, Sonnet 4.6 (1M context)
Context Discipline	Manage your 200k token budget
Learning Log	Auto-document insights
Orchestration	Command > Agent > Skill wiring for complex features
Multi-Phase Dev	Research > Plan > Implement > Review & Commit with validation gates

Installation

Cursor (Recommended)

/add-plugin pro-workflow

The plugin includes 11 skills, 5 agents, and 6 rules that load automatically.

Claude Code — One-Click Plugin Install

/plugin marketplace add rohitg00/pro-workflow
/plugin install pro-workflow@pro-workflow

Claude Code — Build with SQLite Support

cd ~/.claude/plugins/*/pro-workflow
npm install && npm run build

Or load directly

claude --plugin-dir /path/to/pro-workflow

Minimal (Just add to CLAUDE.md)

## Pro Workflow

### Self-Correction
When corrected, propose rule -> add to LEARNED after approval.

### Planning
Multi-file: plan first, wait for "proceed".

### Quality
After edits: lint, typecheck, test.

### LEARNED

Reference Guides

Guide	What's Covered
docs/settings-guide.md	All settings keys, permission modes, hierarchy, sandbox, MCP approval
docs/cli-cheatsheet.md	Every CLI flag, keyboard shortcut, slash command, scripting patterns
docs/orchestration-patterns.md	Command > Agent > Skill architecture, frontmatter reference, agent teams
docs/context-loading.md	CLAUDE.md loading in monorepos, agent memory, skills discovery
docs/cross-agent-workflows.md	Using Claude Code + Cursor together, config mapping, SkillKit translation
docs/new-features.md	Voice mode, agent teams, checkpointing, remote control, new hook events
docs/daily-habits.md	Session habits, debugging tips, terminal setup, anti-patterns

Skills

Skill	Description
pro-workflow	Core workflow patterns for AI-assisted coding
smart-commit	Quality gates, staged review, and conventional commits
wrap-up	End-of-session ritual with change audit and learning capture
learn-rule	Capture corrections as persistent learning rules
parallel-worktrees	Set up git worktrees for zero dead time
replay-learnings	Surface past learnings relevant to the current task
session-handoff	Generate handoff documents for session continuity
insights	Session analytics, correction trends, and productivity metrics
deslop	Remove AI-generated code slop and clean up style
context-optimizer	Token management, context budget, MCP audit
orchestrate	Wire Commands, Agents, and Skills for multi-phase development

Agents

Agent	Purpose	Key Feature
planner	Break down complex tasks	Read-only, approval-gated
reviewer	Code review, security audit	Checklist-based, severity levels
scout	Confidence-gated exploration	Background, worktree-isolated
orchestrator	Multi-phase feature development	Research > Plan > Implement > Review, memory-enabled
debugger	Systematic bug investigation	Hypothesis-driven, root cause analysis

Commands (Claude Code)

When installed as a plugin, commands are available as /pro-workflow:<name>. Short forms like /develop also work directly after plugin install.

Command	Alias	Purpose
/pro-workflow:wrap-up	/wrap-up	End-of-session checklist
/pro-workflow:learn-rule	/learn-rule	Extract correction to memory
/pro-workflow:commit	/commit	Smart commit with quality gates
/pro-workflow:develop	/develop	Multi-phase feature build (Research > Plan > Implement > Review & Commit)
/pro-workflow:doctor	/doctor	Health check for setup and configuration
/pro-workflow:insights	/insights	Session analytics and correction heatmap
/pro-workflow:replay	/replay	Surface past learnings for current task
/pro-workflow:handoff	/handoff	Generate session handoff document
/pro-workflow:search	/search	Search learnings by keyword
/pro-workflow:list	/list	List all stored learnings

Hooks (18 Events)

Hook	When	What
PreToolUse	Before edits	Track edit count, quality gate reminders
PreToolUse	Before git commit/push	Remind about quality gates, wrap-up
PostToolUse	After code edits	Check for console.log, TODOs, secrets
PostToolUse	After tests	Suggest [LEARN] from failures
UserPromptSubmit	Each prompt	Drift detection, correction tracking
SessionStart	New session	Load learnings from database
Stop	Each response	Context-aware reminders
SessionEnd	Session close	Save session stats to database
PreCompact	Before compaction	Save context state
ConfigChange	Settings modified	Detect mid-session changes
Notification	System events	Log permission requests
SubagentStart	Subagent spawns	Log agent lifecycle
SubagentStop	Subagent finishes	Collect results
TaskCompleted	Task marked done	Quality gate on completion
PermissionRequest	Permission dialog	Flag dangerous operations
PostToolUseFailure	Tool fails	Track failures, suggest learnings
TeammateIdle	Teammate goes idle	Detect blockers in agent teams
Setup	Initial setup	One-time initialization (available, not configured)
WorktreeCreate	Worktree created	Set up worktree config (available, not configured)
WorktreeRemove	Worktree removed	Cleanup (available, not configured)

Hook Lifecycle

graph LR
    A[SessionStart] --> B[UserPromptSubmit]
    B --> C[PreToolUse]
    C --> D{Tool Runs}
    D -->|Success| E[PostToolUse]
    D -->|Failure| F[PostToolUseFailure]
    E --> G[Stop]
    F --> G
    G -->|More prompts| B
    G -->|Session ends| H[SessionEnd]

    C -->|Spawns agent| I[SubagentStart]
    I --> J[SubagentStop]

    C -->|Needs permission| K[PermissionRequest]

    style A fill:#6366f1,color:#fff,stroke:none
    style H fill:#ef4444,color:#fff,stroke:none
    style D fill:#f59e0b,color:#000,stroke:none
    style E fill:#22c55e,color:#fff,stroke:none
    style F fill:#ef4444,color:#fff,stroke:none

Rules (Cursor)

Rule	Applies To	Description
quality-gates	Always	Lint, typecheck, and test before commits
atomic-commits	Always	Conventional format, feature branches, specific staging
context-discipline	Always	Read before edit, plan before multi-file changes
self-correction	Always	Capture mistakes as compounding learnings
no-debug-statements	*.{ts,tsx,js,jsx,py,go,rs}	Remove console.log, debugger, print before committing
communication-style	Always	Concise, action-oriented, no over-engineering

Configuration

Settings Example

See settings.example.json for a production-ready configuration including:

• Permission rules (allow/deny/ask)
• Output style and status line
• Auto-compaction at 80%
• Custom spinner verbs and tips
• Attribution disabled
• Plans directory

MCP Config

See mcp-config.example.json for curated server recommendations:

• context7 — Live documentation lookup
• playwright — Browser automation (most token-efficient)
• GitHub — PRs, issues, code search

Rule: Start with 3 MCPs. Add only for concrete needs.

Cross-Agent Support

Works across 32+ AI coding agents via SkillKit:

graph TD
    PW[Pro Workflow] --> CC[Claude Code]
    PW --> CU[Cursor]
    PW --> SK[SkillKit]

    CC --> |"CLAUDE.md + hooks + agents"| CC
    CU --> |".cursorrules + skills + agents"| CU

    SK --> CO[Codex CLI]
    SK --> GC[Gemini CLI]
    SK --> WS[Windsurf]
    SK --> MORE[27+ more]

    style PW fill:#6366f1,color:#fff,stroke:none
    style CC fill:#a78bfa,color:#fff,stroke:none
    style CU fill:#f59e0b,color:#000,stroke:none
    style SK fill:#22c55e,color:#fff,stroke:none

npx skillkit install pro-workflow
npx skillkit translate pro-workflow --agent cursor
npx skillkit translate pro-workflow --agent codex
npx skillkit translate pro-workflow --agent gemini-cli

See docs/cross-agent-workflows.md for configuration mapping between Claude Code, Cursor, Codex, and Gemini CLI.

Structure

pro-workflow/
├── .claude-plugin/              # Claude Code plugin
├── .cursor-plugin/              # Cursor plugin
├── skills/                      # 11 skills (Cursor + Claude Code)
│   ├── pro-workflow/SKILL.md
│   ├── smart-commit/SKILL.md
│   ├── context-optimizer/SKILL.md
│   ├── orchestrate/SKILL.md
│   └── ...
├── agents/                      # 5 agents
│   ├── planner.md
│   ├── reviewer.md
│   ├── scout.md
│   ├── orchestrator.md
│   └── debugger.md
├── commands/                    # 10 slash commands
│   ├── develop.md               # Multi-phase feature build
│   ├── doctor.md                # Health check
│   ├── commit.md
│   └── ...
├── hooks/                       # 18 hook events
│   └── hooks.json
├── docs/                        # Reference guides
│   ├── settings-guide.md
│   ├── cli-cheatsheet.md
│   ├── orchestration-patterns.md
│   ├── context-loading.md
│   ├── cross-agent-workflows.md
│   ├── new-features.md
│   └── daily-habits.md
├── rules/                       # 6 Cursor rules
├── contexts/                    # 3 context modes
├── templates/                   # Split CLAUDE.md templates
├── references/                  # Claude Code resources
├── scripts/                     # Hook scripts
├── settings.example.json        # Production settings template
├── mcp-config.example.json      # Curated MCP recommendations
├── src/                         # TypeScript source (SQLite)
└── config.json

Database Features

Learnings stored in SQLite with FTS5 full-text search:

~/.pro-workflow/
└── data.db    # SQLite database with learnings and sessions

/search testing           # Find all testing-related learnings
/search "file paths"      # Exact phrase search

Philosophy

• Compound improvements - Small corrections lead to big gains over time
• Trust but verify - Let AI work, review at checkpoints
• Zero dead time - Parallel sessions keep momentum
• Memory is precious - Both yours and Claude's
• Orchestrate, don't micromanage - Wire patterns together, let agents execute

Quick Start (30 seconds)

# Option A: Plugin install (Claude Code)
/plugin marketplace add rohitg00/pro-workflow
/plugin install pro-workflow@pro-workflow

# Option B: Plugin install (Cursor)
/add-plugin pro-workflow

# Option C: Manual (any agent)
git clone https://github.com/rohitg00/pro-workflow.git /tmp/pw
cp -r /tmp/pw/templates/split-claude-md/* ./.claude/

# Then try it (use /pro-workflow:<name> with plugin install):
/pro-workflow:develop add user authentication   # Multi-phase feature build
/pro-workflow:doctor                             # Check your setup
/pro-workflow:wrap-up                            # End session properly

If you find this useful, star the repo to help others discover it.



SkillKit Marketplace • Report Issues • Reference Guides

_{Complete AI coding workflow system from production use across Claude Code, Cursor, and beyond.}