Agent Context Manager

Automatically load AGENTS.md files alongside CLAUDE.md for specialized agent-specific instructions.

[ ]

Problem This Solves

Before: Claude Code only reads CLAUDE.md automatically. If you want agent-specific instructions separate from general project context, you have to manually manage them or pollute CLAUDE.md with agent-only rules.

After: Create AGENTS.md in any directory, and Claude Code will automatically detect and load it alongside CLAUDE.md, enabling specialized agent behaviors without manual intervention.

Quick Start

Installation

/plugin install agent-context-manager@claude-code-plugins-plus

Basic Usage

1. Create AGENTS.md in your project root:

cd /path/to/your/project
nano AGENTS.md

1. Add agent-specific rules:

# AGENTS.md - Agent-Specific Instructions

## Agent Behavior Rules

When working with Agent Skills in this project:

1. **Always use TypeScript strict mode** for all generated code
2. **Never create files** without explicit user permission
3. **Follow naming convention**: use kebab-case for all file names
4. **Auto-commit after changes**: Create git commits automatically

## Specialized Workflows

### Code Generation
- Use templates from `./templates/` directory
- Run ESLint after generating any .ts/.js files
- Add comprehensive JSDoc comments

1. That's it! Start Claude Code and the plugin will automatically:

• Detect AGENTS.md
• Load the content
• Apply the rules for your session

Features

⚡ Automatic Loading (Layer 1: Proactive Skill)

The plugin automatically detects and loads AGENTS.md when:

• Starting a new Claude Code session
• Changing directories (via cd command)
• Invoking any other agent skill

User Experience:

📋 Loaded agent-specific context from AGENTS.md

Following specialized agent rules for this session:
- Always use TypeScript strict mode
- Never create files without permission
- Follow kebab-case naming convention

No user action required!

🔔 Directory Change Detection (Layer 2: Hooks)

When you enter a directory with AGENTS.md, you'll see:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📋 AGENTS.md detected in current directory
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

⚡ Agent Context Manager is active

The agent-context-loader skill will automatically load
agent-specific instructions from AGENTS.md

Location: /path/to/project/AGENTS.md

What happens next:
  1. Claude will read AGENTS.md automatically
  2. Agent-specific rules will be incorporated
  3. Instructions will be active for this session

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🔄 Manual Synchronization (Layer 3: Slash Command)

Permanently merge AGENTS.md into CLAUDE.md:

/sync-agent-context

What it does:

1. Finds all AGENTS.md files in your project
2. Reads their content
3. Merges into CLAUDE.md under "## Agent-Specific Instructions" section
4. Creates backup: CLAUDE.md.backup.TIMESTAMP

Result:

## Agent-Specific Instructions

### Root Directory Agent Rules (./AGENTS.md)
[Content from ./AGENTS.md]

### Subproject Agent Rules (./packages/app/AGENTS.md)
[Content from ./packages/app/AGENTS.md]

Architecture

Three-Layer Redundancy System

Layer 1: Proactive Skill (agent-context-loader)
   ↓
   Automatically invoked when context needed
   ↓
   Reads AGENTS.md using Read tool
   ↓
   Loads into session context

Layer 2: Directory Change Hook (check-agents-md.sh)
   ↓
   Runs on cd, session start
   ↓
   Prints detection message
   ↓
   Reminds Claude to load AGENTS.md

Layer 3: Manual Sync Command (/sync-agent-context)
   ↓
…

Why three layers?

1. Layer 1 = Ideal (fully automatic)
2. Layer 2 = Backup (semi-automatic reminder)
3. Layer 3 = Fallback (manual permanent merge)

This ensures AGENTS.md is always loaded, even if one layer fails.

Usage Examples

Example 1: Single Project with Agent Rules

Project Structure:

my-project/
├── CLAUDE.md           # General project context
├── AGENTS.md           # Agent-specific rules
└── src/
    └── index.ts

Workflow:

cd my-project
# Plugin automatically detects AGENTS.md
# Claude loads both CLAUDE.md and AGENTS.md
# Agent rules are active for this session

Example 2: Multi-Package Monorepo

Project Structure:

monorepo/
├── CLAUDE.md
├── AGENTS.md                   # Root-level agent rules
└── packages/
    ├── app/
    │   └── AGENTS.md           # App-specific agent rules
    └── api/
        └── AGENTS.md           # API-specific agent rules