Getting Started

Thank you for your interest in contributing to the rlm-rs Claude Code plugin.

Getting Started

Prerequisites

1. Claude Code with plugin support
2. rlm-rs CLI installed:

   cargo install --git https://github.com/zircote/rlm rlm-rs

3. Familiarity with Claude Code plugin structure

Development Setup

1. Clone the repository:

   git clone https://github.com/zircote/rlm-rs-plugin.git
   cd rlm-rs-plugin

2. Run Claude Code with the plugin directory:

   claude --plugin-dir .

3. Verify loading by running /help - you should see rlm-rs:rlm-init and other commands.

Project Structure

rlm-rs-plugin/
├── .claude-plugin/
│   └── plugin.json       # Plugin manifest
├── agents/
│   ├── rlm-subcall.md    # Chunk analysis agent
│   └── rlm-synthesizer.md # Result synthesis agent
├── commands/
│   ├── rlm-init.md       # Initialize command
│   ├── rlm-load.md       # Load file command
│   ├── rlm-status.md     # Status command
│   └── rlm-query.md      # Query command
├── skills/
│   ├── rlm/
│   │   ├── SKILL.md      # Main orchestration skill
│   │   └── references/
│   │       └── cli-reference.md
│   └── rlm-chunking/
│       └── SKILL.md      # Chunking strategy guide
├── hooks/                # (future) Event hooks
├── docs/
│   ├── ARCHITECTURE.md       # Architecture documentation
│   └── TECHNICAL-OVERVIEW.md # Technical deep-dive
├── README.md
└── CONTRIBUTING.md

Component Guidelines

Skills

Skills are proactive capabilities triggered by user intent.

File location: skills/<skill-name>/SKILL.md

Frontmatter format:

---
name: skill-name
description: Trigger phrases and when to use this skill
version: 0.1.0
allowed-tools:
  - Read
  - Write
  - Bash
  - Task
---

Best practices:

• Include clear trigger phrases in description
• Document prerequisites and dependencies
• Provide step-by-step workflow instructions
• Include example commands and outputs

Commands

Commands are explicit slash commands invoked by users.

File location: commands/<command-name>.md

Frontmatter format:

---
name: command-name
description: Brief description
argument-hint: arg1=<value> [arg2=<optional>]
arguments:
  - name: arg1
    description: What this argument does
    required: true
  - name: arg2
    description: Optional argument
    required: false
---

Best practices:

• Keep commands focused on single actions
• Validate inputs in the bash script
• Provide helpful error messages
• Show status after completion

Agents

Agents are subprocesses invoked via the Task tool.

File location: agents/<agent-name>.md

Frontmatter format:

---
name: agent-name
description: When this agent is invoked
model: haiku|sonnet|opus
tools:
  - Read
  - Bash
color: cyan|green|blue|etc
---

Best practices:

• Choose appropriate model for the task
• Define clear input/output contracts
• Keep agent scope narrow and focused
• Document constraints and limitations

Testing Changes

Manual Testing

1. Make changes to plugin files
2. Restart Claude Code or reload plugins
3. Test the modified component:

   # Test a command
   /rlm-init
   
   # Test a skill by describing intent
   "I need to process a large log file using RLM"
   
   # Test an agent via Task tool (pass chunk ID from search results)
   "Analyze chunk ID 42 from the loaded buffer"

Test Scenarios

Basic workflow:

# 1. Create test file
echo "Test content for RLM processing" > test.txt

# 2. Initialize
/rlm-init

# 3. Load
/rlm-load file=test.txt

# 4. Check status
/rlm-status

Chunking strategies:

# Test semantic chunking with markdown
/rlm-load file=README.md chunker=semantic

# Test fixed chunking with overlap
/rlm-load file=test.log chunker=fixed chunk-size=1000 overlap=100

Submitting Changes

Pull Request Process

1. Fork the repository
2. Create a feature branch:

   git checkout -b feature/my-improvement

3. Make your changes
4. Test thoroughly
5. Commit with descriptive message:

   git commit -m "Add parallel chunk processing to rlm skill"

6. Push and create PR:

   git push origin feature/my-improvement

PR Guidelines

• Title: Clear, concise description of change
• Description: Include:
  • What changed and why
  • How to test the change
  • Any breaking changes
• Size: Keep PRs focused; split large changes

Commit Messages

Follow conventional commits:

• feat: New feature
• fix: Bug fix
• docs: Documentation only
• refactor: Code change that neither fixes nor adds
• test: Adding tests
• chore: Maintenance tasks

Code Style

Markdown

• Use ATX-style headers (#, ##, ###)
• Include code fence language identifiers
• Keep lines under 100 characters where practical
• Use reference-style links for repeated URLs

YAML Frontmatter

• Use lowercase property names
• Quote strings with special characters
• Keep descriptions concise but complete

Bash Scripts

• Always check for prerequisites
• Provide clear error messages
• Use set -e for strict error handling (when appropriate)
• Quote variables: "${variable}"

Getting Help

• Issues: GitHub Issues
• Discussions: GitHub Discussions

License

By contributing, you agree that your contributions will be licensed under the MIT License.