agent-workspace-manager（Agent 工作区经理）

Agent Workspace Manager

Overview

This skill provides a systematic approach to organizing and maintaining agent workspace directories. It ensures documents are placed in appropriate functional directories, detects structure issues, and maintains content freshness.

Core Capabilities

1. Structure Initialization

Initialize a standardized workspace structure for new agents:

workspace/
├── MEMORY.md              # Core memory index (agent reads this first)
├── memory/                # Time-based memory fragments
├── knowledge/             # Stable domain knowledge
│   └── sql/               # Core business SQL (NEW)
├── tasks/                 # Temporary/project-specific items
├── decisions/             # Decision records with rationale
├── preferences/           # User preferences and settings
├── archives/              # Expired but preserved content
│   └── sql_versions/      # Historical SQL versions (NEW)
├── templates/             # Reusable templates
│   └── sql/               # Reusable SQL templates (NEW)
└── scripts/               # Scripts and utilities (NEW)

Usage: When setting up a new workspace, create this structure to enable efficient information retrieval.

2. Document Classification

Every document must be placed in a directory matching its functional purpose:

Classification Rule: Before creating any document, determine its primary function. Place it in the matching directory. Never place files in incorrect directories.

3. Code File Classification (NEW)

Code files (SQL, Python, Shell, etc.) must follow these rules:

SQL File Naming Rules:

• ✅ DO: Use semantic names: 神策宽表转JSON.sql
• ❌ DON'T: Use version suffixes: 神策宽表转JSON_v1.sql, 神策宽表转JSON_最终版.sql
• ❌ DON'T: Use functional suffixes: 神策宽表转JSON_修正版.sql, 神策宽表转JSON_完整版.sql

When to Create New Version:

• If SQL needs update, modify the existing file
• Archive old version to archives/sql_versions/ with timestamp if needed
• Keep only ONE active version in knowledge/sql/

4. Workspace Health Check

Run health checks to identify issues:

• Misplaced files: Documents in wrong directories
• Duplicates: Similar content in multiple files
• Code file duplicates: Multiple versions of same SQL/script (NEW)
• Stale content: Outdated information that needs refresh
• Orphan files: Files without clear purpose or references
• Naming violations: Files with version suffixes (NEW)

Output: A report with specific file recommendations.

5. Duplicate File Detection (ENHANCED)

Detect duplicate files using these methods:

For Document Files (.md)

1. Name similarity: Files with similar names (e.g., report-v1.md and report-v2.md)
2. Content similarity: Files with >80% content overlap
3. Functional similarity: Files serving the same purpose

For Code Files (.sql, .py, .sh)

1. Version suffixes: Files with _v1, _v2, _final, _修正版, _完整版 etc.
2. Same functionality: Multiple SQL files for the same transformation/query
3. Timestamp in name: Files with timestamps that could be archived

Detection Command:

# Find SQL files with version suffixes
find . -name "*.sql" -type f | grep -E "(_v[0-9]|_最终|_完整|_修正|_修复)"

# Find duplicate MD files by name pattern
find . -name "*.md" -type f | grep -E "(_v[0-9]|_最终|_完整)"

6. Content Freshness Maintenance

Maintain document freshness through:

1. Merge: Combine duplicate or similar files, preserving all key information
2. Archive: Move expired content to archives/ with timestamp
3. Refresh: Update stale knowledge with latest information
4. Prune: Remove truly redundant content (only when no key info is lost)
5. Consolidate: Merge multiple versions into one, archive old versions (NEW)

Safety Principle: Never delete content that might contain unique valuable information. Archive instead.

7. Structure Extension Detection

When existing directories cannot accommodate new functional needs:

1. Detect: Identify documents that don't fit current categories
2. Propose: Suggest new directories or structure reorganization
3. Implement: Create new directories following naming conventions
4. Document: Update this skill's structure reference

Example: If knowledge/ grows too large with mixed domain types, split into knowledge/product/, knowledge/technical/, knowledge/business/.

Workflow

For New Workspace Setup

1. Determine agent type (personal assistant, project manager, knowledge base, etc.)
2. See references/structure-templates.md for appropriate template
3. Create directories and initial files
4. Set up MEMORY.md as the entry point

For Workspace Audit

1. Scan all files in workspace
2. Classify each file by function
3. Check for misplaced files
4. Detect duplicate files (including code files)
5. Generate health report
6. Propose fixes for user approval

For Content Cleanup

1. Identify duplicates/similar files
2. Detect version-suffixed files (NEW)
3. Propose merge candidates
4. Identify expired content
5. Archive with preservation
6. Verify no key information is lost

For Code File Cleanup (NEW)

1. Scan all code files (.sql, .py, .sh)
2. Identify files with version suffixes
3. Determine the latest/best version
4. Keep latest in appropriate directory (knowledge/sql/ or templates/sql/)
5. Archive old versions to archives/sql_versions/
6. Remove version suffixes from filenames

Resources

references/

• structure-templates.md - Pre-defined workspace templates for different agent types
• classification-rules.md - Detailed rules for document classification (UPDATED)

Multi-Agent Sharing

This skill is designed to be copyable across agents:

1. Install the same skill in multiple agent workspaces
2. Each agent follows the same structure conventions
3. Shared knowledge can be exchanged between agents
4. Consistent structure enables cross-agent collaboration

Changelog

v1.1 (2026-03-25)

• Added code file classification rules (SQL, Python, Shell)
• Added duplicate file detection methods
• Added file naming best practices
• Enhanced workspace health check to detect version-suffixed files
• Added SQL-specific cleanup workflow