agents-md-pro
Scannednpx machina-cli add skill Melon4Program/ai-skills/agents-md-pro --openclawAGENTS.md Pro
Create token-efficient AGENTS.md files that maximize clarity with minimal tokens.
Core Principles
- Token efficiency - Every word justifies its cost
- Commands over explanations - Show, don't tell
- Reference configs - Point to
.eslintrc, never duplicate - Model-agnostic - Universal terminology only
- Condensed default - Always minimal output
Input
Required: Project directory path If missing: Request from user
Workflow Router
Map user request to workflow:
- Create → workflows.md
- Optimize/condense → workflows.md
- Update/refresh → workflows.md
- Validate → workflows.md
Quick Reference
Output template - Standard repo:
# [Project] | [Tech Stack]
## COMMANDS
- Dev: `cmd` | Build: `cmd` | Test: `cmd` | Lint: `cmd --fix`
## STRUCTURE
- `dir/` - purpose
## PATTERNS
[1-2 key patterns with minimal code]
## CODE STYLE
See `.eslintrc`, `.prettierrc`
## DOMAIN
| Term | Definition |
## SECURITY
[Auth/validation only]
## GIT
Format: `convention`
Line limits:
- Standard: ≤150 lines
- Monorepo root: ≤50 lines
- Sub-project: ≤100 lines
Target tokens:
- Standard: 500-800
- Monorepo root: 300-400
- Sub-project: 400-600
Resources
Load as needed:
- Workflows: workflows.md - All 4 workflows with step-by-step procedures
- Optimization: optimization-patterns.md - Token reduction techniques
- Validation: validation-rules.md - Quality checklist and scoring
- Anti-patterns: anti-patterns.md - Common bloat patterns to avoid
Source
git clone https://github.com/Melon4Program/ai-skills/blob/main/skills/agents-md-pro/SKILL.mdView on GitHub Overview
AGENTS.md Pro creates, optimizes, updates, and validates AGENTS.md files with maximum token efficiency. It delivers condensed, actionable commands and references to configs in a model-agnostic style. This helps maintain clear AI agent documentation across repositories with minimal verbosity.
How This Skill Works
When a user request is received, the tool routes it to the appropriate workflow (Create, Optimize/condense, Update/refresh, Validate) via the Workflow Router and produces a standard, token-conscious output template. It emphasizes referencing existing configs (like .eslintrc) rather than duplicating them, and keeps output within strict line/token targets for easy parsing by AI agents.
When to Use It
- Create new AGENTS.md files for any repository
- Optimize or condense existing AGENTS.md to reduce token count
- Update or refresh AGENTS.md to sync with codebase changes
- Validate AGENTS.md quality and completeness
- Improve AGENTS.md to be more effective for AI agents
Quick Start
- Step 1: Provide the project directory path to initiate the workflow
- Step 2: Choose the workflow: Create, Optimize/condense, Update/refresh, or Validate
- Step 3: Review the generated AGENTS.md snippet and commit to the repo
Best Practices
- Reference configs (e.g., .eslintrc, .prettierrc) instead of duplicating rules
- Keep language model-agnostic with universal terminology
- Aim for concise sections: COMMANDS, STRUCTURE, PATTERNS, CODE STYLE, GIT
- Target token counts: Standard 500-800 tokens; adjust for repo size
- Validate against established rules and anti-patterns before finalizing
Example Use Cases
- New repo: generate a compact AGENTS.md with PROJECT + TECH STACK, essential COMMANDS, and quick patterns
- Existing AGENTS.md compressed from 1200 tokens to 450 tokens without losing meaning
- Sync AGENTS.md with the latest code changes after a major refactor
- Run validation to ensure completeness of DOMAIN, SECURITY, and GIT sections
- Improve AGENTS.md to emphasize actionable commands over long explanations
