Intelligent agent context file editing for OpenClaw

A skill that detects when you want to modify agent behavior and guides safe, structured edits to core context files like AGENTS.md, SOUL.md, IDENTITY.md, and others.

• Auto-detects configuration change intent from natural language
• Routes changes to the correct context file (operational vs personality vs tools)
• Prevents duplication by checking existing content before adding
• Enforces size limits (20K char per file) with proactive warnings
• Validates format against Claude prompt engineering best practices
• Provides rollback guidance when changes don't work as expected

Instead of manually editing context files and hoping for the best, this skill walks through a structured workflow: identify target file → check current state → validate before applying → verify after changes.

Install from ClawHub:

clawhub install agent-config

Or clone manually into your OpenClaw skills directory:

cd ~/clawd/skills/
git clone https://github.com/ThatGuySizemore/agent-config.git

agent-config/
├── SKILL.md                    # Main workflow and usage guide
└── references/
    ├── file-map.md            # What each context file does, decision trees
    ├── claude-patterns.md     # Instruction formats that work for Claude
    └── change-protocol.md     # Detailed change process and validation

The skill triggers automatically when conversation involves:

• Changing agent behavior or rules
• Updating personality or tone
• Modifying operational procedures
• Adding safety rules or constraints
• Refining delegation patterns
• Adjusting memory workflows
• Updating tool conventions

Just describe what you want to change naturally, and the agent will use this skill to apply it correctly.

• ClawHub page
• Designed for OpenClaw's context file architecture
• Works with both main agent sessions and subagent contexts

MIT License - Copyright 2026 Josh Sizemore