Skip to content

README.en.md

Latest commit

 

History

History
127 lines (92 loc) · 4.97 KB

README.en.md

File metadata and controls

127 lines (92 loc) · 4.97 KB

Master Control Expert (MCE)

A structured task processing framework for Claude Code. MCE transforms any request into executable, governed workflows.

Overview

Master Control Expert (MCE) is a task classification and execution framework that helps Claude Code efficiently handle various types of tasks. Through standardized processes and prompt templates, MCE ensures every task receives systematic, high-quality processing.

Core Workflow

Receive Task → Intake → Classify → Plan → Deliver → Evolve

Five-Step Method

  1. Intake - Collect necessary information using the universal checklist
  2. Classify - Categorize the task into A-H categories
  3. Control - Output control summary and definition of done
  4. Execute - Apply corresponding prompt templates and produce concrete deliverables
  5. Evolve - Self-assess and propose category additions or prompt improvements

Task Categories (A-H)

Category Name Use Cases
A Logic & Argumentation Argument audits, validity checks, counterexamples, "cold-water reviews"
B Math / Theory Explanation Tiered explanations, concept maps, learning routes, book lists
C Engineering Delivery Systems, refactors, specs, incidents, deployment
D Product / Mechanism Design Governance, rules, incentives, anti-abuse, sustainability
E Writing & Publishing Outlines, rewrites, de-duplication, editorial guidance
F Policy / Legal / Compliance Legal reasoning, compliance checklists, policy analysis
G Research Curation / Study Guide Knowledge maps, reading guides, tagging systems, bibliographies
H Events & Project Management Conferences, forums, awards, competitions, dinners

File Structure

master-control-expert/
├── SKILL.md                    # Skill definition file (for Claude Code)
├── README.md                   # This file (Chinese)
├── README.en.md                # English version
└── playbook/                   # Playbook directory
    ├── intake_checklist.md     # Universal intake checklist
    ├── prompt_catalog.md       # A-H category prompt skeletons (with version & metrics)
    ├── control_plan.md         # Control plan template
    ├── decision_log.md         # Decision log
    ├── runbook.md              # Runbook
    ├── status_report.md        # Status report template
    └── evolution_protocol.md   # Self-evolution protocol (new categories / prompt improvements)

Usage

As a Claude Code Skill

When activated in Claude Code, the system automatically follows the MCE framework to process tasks:

/master-control-expert

Manual Reference

  1. First consult playbook/intake_checklist.md for intake
  2. Based on task type, consult corresponding template in playbook/prompt_catalog.md
  3. Use playbook/control_plan.md to plan execution
  4. Record important decisions in playbook/decision_log.md

Key Principles

  • Direct and Operational: Provide clear checklists, tables, and numbered steps
  • No Stalling: Deliver immediately, don't promise future work
  • Explicit Assumptions: When information is insufficient, make reasonable assumptions and label them
  • Track Decisions: Explicitly record decisions, maintain a decision log
  • Concrete Deliverables: Files, content blocks, executable steps
  • Self-Evolving: Continuously assess category coverage and prompt effectiveness, proactively improve

Self-Evolution Features

MCE has self-evolution capabilities and can automatically optimize based on usage:

New Category Workflow

When a task doesn't fit any A-H category, MCE will:

  1. Assess whether a new category is needed
  2. Present a new category proposal to the user (name, definition, prompt skeleton)
  3. Upon user confirmation, automatically update relevant files
  4. Record in evolution_protocol.md

Triggers:

  • Task differs from A-H by > 70%
  • 3 consecutive tasks fall into Other with related themes
  • User explicitly requests a new category

Prompt Improvement Workflow

When a category's prompt effectiveness is suboptimal, MCE will:

  1. Diagnose the issue (missing/redundant/misordered)
  2. Present improvement proposal to user (before/after comparison)
  3. Upon user confirmation, update prompt_catalog.md
  4. Increment version number and record improvement log

Triggers:

  • User feedback indicates poor effectiveness
  • Prompt skeleton found to miss key items
  • User repeatedly overrides default sections
  • Same category requires major adjustments 2+ times

See playbook/evolution_protocol.md for details.

Sample Outputs

For each task, MCE produces:

  • Control Summary: Objective, deliverables, constraints, assumptions, risks, definition of done
  • Execution Plan: Milestones, resources, steps
  • Concrete Deliverables: Actual files/content blocks
  • Decision Log Entries: Record key decisions

Contributing

Improvements to this framework are welcome via Issues or PRs.

License

MIT License