workspace-governance

中文文档

A methodology-first workspace governance skill for AI agents.
The goal is safe, reversible, and traceable workspace management without forcing a fixed folder scaffold.

Document Metadata

• Author: Mars
• GitHub: Mars2003
• Date: 2026-04-28

Core Principles (Short)

• Boundary before structure
• Plan before action
• Reversible before optimized
• Explicit confirmation for delete and bulk move
• Protect sensitive files and VCS metadata by default

Quick Start

1) Install the skill

# Claude Code (global)
mkdir -p ~/.claude/skills/workspace-governance
cp SKILL.md ~/.claude/skills/workspace-governance/

# Cursor (project-level)
mkdir -p .cursor/skills/workspace-governance
cp SKILL.md .cursor/skills/workspace-governance/

2) Project-level adaptation (optional but recommended)

This repository includes SKILL_ADAPT.yaml for boundary, protection, batch execution, and logging preferences.
The skill works without it, but with it, behavior is more stable and consistent across sessions.

3) Recommended trigger intents

• "organize workspace"
• "audit first, then cleanup"
• "archive project xxx"
• "create project yyy with boundaries"

4) One-command validation (recommended)

make check

Suggested Directories

Beyond SKILL.md, you can optionally maintain:

• references/: manuals, process notes, strategy examples
• assets/: diagrams, screenshots, and visual resources
• scripts/: helper scripts for checks, batch operations, exports
• tools/: tool configs and integration notes

Detailed Docs

• English governance manual (detailed): references/Governance-Manual.md
• Chinese governance manual (detailed): references/治理手册.zh-CN.md
• Minimal run example: examples/minimal-governance-run.md

Project Meta

• Changelog: CHANGELOG.md
• Contributing guide: CONTRIBUTING.md

License

MIT