agents-md-improver

Propose cohesive patches to AGENTS.md from user corrections.

38|5|Updated Dec 20, 2025

Search Tags:#automation #workflow #patch #consolidation #AGENTS.md #version-control #agent-instructions

Ready to run

npx skills add https://github.com/ericmjl/skills --skill agents-md-improver-ericmjl

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agents-md-improver
Source: https://github.com/ericmjl/skills/tree/main/skills/agents-md-improver
Command: npx skills add https://github.com/ericmjl/skills --skill agents-md-improver-ericmjl

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Keeps repo-local agent instructions aligned by proposing integrated edits to AGENTS.md whenever a user corrects the agent's behavior or requests changes to AGENTS.md, CLAUDE.md, .claude/CLAUDE.md, or GEMINI.md.

Core Features & Use Cases

Propose durable patches to AGENTS.md that incorporate user corrections in a cohesive, conflict-free manner.
Detect and resolve contradictions between instructions, and coordinate consolidation when per-instruction files exist (CLAUDE.md and related files).
Guide follow-up actions such as deleting, symlinking, or stubbing CLAUDE files to point to AGENTS.md, maintaining a single canonical source of truth.

Quick Start

Draft a patch to AGENTS.md that incorporates the user's correction and present it for review.

Frequently Asked Questions about agents-md-improver

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I keep AGENTS.md and CLAUDE.md instruction files in sync?▼

You can consolidate repo-local agent instructions by drafting cohesive patches to AGENTS.md that incorporate user corrections and resolve contradictions across CLAUDE.md, .claude/CLAUDE.md, and GEMINI.md files.

How do I resolve contradictions between CLAUDE.md and GEMINI.md?▼

Resolving contradictions between instruction files involves drafting a cohesive patch to AGENTS.md and following a conflict resolution workflow to establish a single canonical source of truth.

How do I consolidate multiple agent instruction files into one source of truth?▼

Consolidating agent instruction files uses a workflow that proposes patches to AGENTS.md and manages follow-ups like deleting, symlinking, or stubbing CLAUDE files to point to AGENTS.md.

Can I automatically update AGENTS.md when correcting agent behavior?▼

Yes, updating AGENTS.md automatically in response to user corrections is possible by drafting durable, conflict-free patches that integrate the behavioral changes directly into the instruction file.

What is the best way to manage repo-local agent instructions across different AI coding tools?▼

Managing repo-local agent instructions across tools is best handled by proposing integrated edits to AGENTS.md and coordinating consolidation of per-instruction files like CLAUDE.md and GEMINI.md.

Do I need separate instruction files for different AI agents in the same repository?▼

You do not need separate instruction files; you can maintain a single canonical source by stubbing or symlinking files like CLAUDE.md to point to AGENTS.md for consistent agent behavior.