maxwell2732/claudecode-research-harness-workflow/workflow-guide

Claude Code and agent skill by maxwell2732/claudecode-research-harness-workflow

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Explicit helper for Cursor PM ↔ Claude Code two-agent workflow guidance. Do NOT load for: solo implementation, workflow setup, handoff execution, or general process coaching.

Install

npx vskill@latest install maxwell2732/claudecode-research-harness-workflow/workflow-guide

SKILL.md

Workflow Guide Skill

A skill providing guidance on the Cursor ↔ Claude Code 2-agent workflow.


Trigger Phrases

This skill is invoked by the following phrases:

  • "Tell me about the workflow"
  • "How do I collaborate with Cursor?"
  • "Explain the work process"
  • "How should I proceed?"
  • "how does the workflow work?"
  • "explain 2-agent workflow"

Overview

This skill explains the role assignment and collaboration method between Cursor (PM) and Claude Code (Worker).


2-Agent Workflow

Role Assignment
AgentRoleResponsibilities
CursorPM (Project Manager)Task assignment, review, production deploy decisions
Claude CodeWorkerImplementation, testing, CI fixes, staging deploy
Workflow Diagram
┌─────────────────────────────────────────────────────────┐
│ Cursor (PM) │
│ - Add tasks to Plans.md │
│ - Delegate work to Claude Code (/handoff-to-claude) │
│ - Review completion reports │
│ - Make production deploy decisions │
└─────────────────────┬───────────────────────────────────┘
│ Task delegation
▼
┌─────────────────────────────────────────────────────────┐
│ Claude Code (Worker) │
│ - Execute tasks with /work (parallel execution) │
│ - Implement → Test → Commit │
│ - Auto-fix CI failures (up to 3 times) │
│ - Completion report with /handoff-to-cursor │
└─────────────────────┬───────────────────────────────────┘
│ Completion report
▼
┌─────────────────────────────────────────────────────────┐
│ Cursor (PM) │
│ - Review changes │
│ - Verify on staging │
│ - Execute production deploy (after approval) │
└─────────────────────────────────────────────────────────┘

Task Management with Plans.md

Marker List
MarkerMeaningSet by
pm:requestedRequested by PM (compatible: cursor:requested)PM (Cursor/PM Claude)
cc:TODOClaude Code not startedEither
cc:WIPClaude Code in progressClaude Code
cc:doneClaude Code completeClaude Code
pm:confirmedPM confirmation complete (compatible: cursor:confirmed)PM (Cursor/PM Claude)
cursor:requested(compatible) Same as pm:requestedCursor
cursor:confirmed(compatible) Same as pm:confirmedCursor
blockedBlockedEither
Task State Transitions
pm:requested → cc:WIP → cc:done → pm:confirmed

Key Commands

Claude Code Side
CommandPurpose
/harness-initProject setup
/plan-with-agentPlanning/task breakdown
/workTask execution (parallel execution supported)
/handoff-to-cursorCompletion report (to Cursor PM)
/sync-statusStatus check
Skills (Auto-invoked in Conversation)
SkillTrigger Example
handoff-to-pm"Report completion to PM"
handoff-to-impl"Hand off to implementation"
Cursor Side (For Reference)
CommandPurpose
/handoff-to-claudeDelegate tasks to Claude Code
/review-cc-workReview completion reports

CI/CD Rules

Claude Code's Scope of Responsibility
  • ✅ Up to staging deploy
  • ✅ Auto-fix CI failures (up to 3 times)
  • ❌ Production deploy is prohibited
3-Strike Rule

If CI fails 3 times in a row:

  1. Stop auto-fix
  2. Generate escalation report
  3. Defer decision to Cursor

Frequently Asked Questions

Q: What if Cursor is not available?

A: Even when working alone, task management with Plans.md is recommended. Perform production deploys manually and carefully.

Q: What if a task is unclear?

A: Request clarification from Cursor, or use /sync-status to organize the current state.

Q: What if CI keeps failing?

A: Do not auto-fix more than 3 times; escalate to Cursor.


Related Documents

  • AGENTS.md - Detailed role assignments
  • CLAUDE.md - Claude Code specific configuration
  • Plans.md - Task management file

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/maxwell2732/claudecode-research-harness-workflow/workflow-guide/badge)](https://verified-skill.com/skills/maxwell2732/claudecode-research-harness-workflow/workflow-guide)

Something wrong with this skill or its page? Tell us.