agent-md-refactor
Refactor bloated AGENTS.md, CLAUDE.md, or similar agent instruction files to follow progressive disclosure principles. Splits monolithic files into organized, linked documentation.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 3.3w
- Forks
- 3.9k
01 前言 Front Matter
- name
- agent-md-refactor
- license
- MIT
- description
- Refactor bloated AGENTS.md, CLAUDE.md, or similar agent instruction files to follow progressive disclosure principles. Splits monolithic files into organized, linked documentation.
02 资源文件 4
- 参考文档agents/openai.yaml282 B
- 参考文档README.md7.2 KB
- 参考文档references/experiment-template.md159 B
- 参考文档references/mvp-patterns.md256 B
03 SKILL.md 正文
Agent MD Refactor
Refactor bloated agent instruction files (AGENTS.md, CLAUDE.md, COPILOT.md, etc.) to follow progressive disclosure principles - keeping essentials at root and organizing the rest into linked, categorized files.
Triggers
Use this skill when:
- "refactor my AGENTS.md" / "refactor my CLAUDE.md"
- "split my agent instructions"
- "organize my CLAUDE.md file"
- "my AGENTS.md is too long"
- "progressive disclosure for my instructions"
- "clean up my agent config"
Quick Reference
| Phase | Action | Output |
|---|---|---|
| 1. Analyze | Find contradictions | List of conflicts to resolve |
| 2. Extract | Identify essentials | Core instructions for root file |
| 3. Categorize | Group remaining instructions | Logical categories |
| 4. Structure | Create file hierarchy | Root + linked files |
| 5. Prune | Flag for deletion | Redundant/vague instructions |
Process
Phase 1: Find Contradictions
Identify any instructions that conflict with each other.
Look for:
- Contradictory style guidelines (e.g., "use semicolons" vs "no semicolons")
- Conflicting workflow instructions
- Incompatible tool preferences
- Mutually exclusive patterns
For each contradiction found:
## Contradiction Found
**Instruction A:** [quote]
**Instruction B:** [quote]
**Question:** Which should take precedence, or should both be conditional?
Ask the user to resolve before proceeding.
Phase 2: Identify the Essentials
Extract ONLY what belongs in the root agent file. The root should be minimal - information that applies to every single task.
Essential content (keep in root):
| Category | Example |
|---|---|
| Project description | One sentence: "A React dashboard for analytics" |
| Package manager | Only if not npm (e.g., "Uses pnpm") |
| Non-standard commands | Custom build/test/typecheck commands |
| Critical overrides | Things that MUST override defaults |
| Universal rules | Applies to 100% of tasks |
NOT essential (move to linked files):
- Language-specific conventions
- Testing guidelines
- Code style details
- Framework patterns
- Documentation standards
- Git workflow details
Phase 3: Group the Rest
Organize remaining instructions into logical categories.
Common categories:
| Category | Contents |
|---|---|
typescript.md |
TS conventions, type patterns, strict mode rules |
testing.md |
Test frameworks, coverage, mocking patterns |
code-style.md |
Formatting, naming, comments, structure |
git-workflow.md |
Commits, branches, PRs, reviews |
architecture.md |
Patterns, folder structure, dependencies |
api-design.md |
REST/GraphQL conventions, error handling |
security.md |
Auth patterns, input validation, secrets |
performance.md |
Optimization rules, caching, lazy loading |
Grouping rules:
- Each file should be self-contained for its topic
- Aim for 3-8 files (not too granular, not too broad)
- Name files clearly:
{topic}.md - Include only actionable instructions
Phase 4: Create the File Structure
Output structure:
project-root/
├── CLAUDE.md (or AGENTS.md) # Minimal root with links
└── .claude/ # Or docs/agent-instructions/
├── typescript.md
├── testing.md
├── code-style.md
├── git-workflow.md
└── architecture.md
Root file template:
# Project Name
One-sentence description of the project.
## Quick Reference
- **Package Manager:** pnpm
- **Build:** `pnpm build`
- **Test:** `pnpm test`
- **Typecheck:** `pnpm typecheck`
## Detailed Instructions
For specific guidelines, see:
- [TypeScript Conventions](.claude/typescript.md)
- [Testing Guidelines](.claude/testing.md)
- [Code Style](.claude/code-style.md)
- [Git Workflow](.claude/git-workflow.md)
- [Architecture Patterns](.claude/architecture.md)
Each linked file template:
# {Topic} Guidelines
## Overview
Brief context for when these guidelines apply.
## Rules
### Rule Category 1
- Specific, actionable instruction
- Another specific instruction
### Rule Category 2
- Specific, actionable instruction
## Examples
### Good
\`\`\`typescript
// Example of correct pattern
\`\`\`
### Avoid
\`\`\`typescript
// Example of what not to do
\`\`\`
Phase 5: Flag for Deletion
Identify instructions that should be removed entirely.
Delete if:
| Criterion | Example | Why Delete |
|---|---|---|
| Redundant | "Use TypeScript" (in a .ts project) | Agent already knows |
| Too vague | "Write clean code" | Not actionable |
| Overly obvious | "Don't introduce bugs" | Wastes context |
| Default behavior | "Use descriptive variable names" | Standard practice |
| Outdated | References deprecated APIs | No longer applies |
Output format:
## Flagged for Deletion
| Instruction | Reason |
|-------------|--------|
| "Write clean, maintainable code" | Too vague to be actionable |
| "Use TypeScript" | Redundant - project is already TS |
| "Don't commit secrets" | Agent already knows this |
| "Follow best practices" | Meaningless without specifics |
Execution Checklist
[ ] Phase 1: All contradictions identified and resolved
[ ] Phase 2: Root file contains ONLY essentials
[ ] Phase 3: All remaining instructions categorized
[ ] Phase 4: File structure created with proper links
[ ] Phase 5: Redundant/vague instructions removed
[ ] Verify: Each linked file is self-contained
[ ] Verify: Root file is under 50 lines
[ ] Verify: All links work correctly
Anti-Patterns
| Avoid | Why | Instead |
|---|---|---|
| Keeping everything in root | Bloated, hard to maintain | Split into linked files |
| Too many categories | Fragmentation | Consolidate related topics |
| Vague instructions | Wastes tokens, no value | Be specific or delete |
| Duplicating defaults | Agent already knows | Only override when needed |
| Deep nesting | Hard to navigate | Flat structure with links |
Examples
Before (Bloated Root)
# CLAUDE.md
This is a React project.
## Code Style
- Use 2 spaces
- Use semicolons
- Prefer const over let
- Use arrow functions
... (200 more lines)
## Testing
- Use Jest
- Coverage > 80%
... (100 more lines)
## TypeScript
- Enable strict mode
... (150 more lines)
After (Progressive Disclosure)
# CLAUDE.md
React dashboard for real-time analytics visualization.
## Commands
- `pnpm dev` - Start development server
- `pnpm test` - Run tests with coverage
- `pnpm build` - Production build
## Guidelines
- [Code Style](.claude/code-style.md)
- [Testing](.claude/testing.md)
- [TypeScript](.claude/typescript.md)
Verification
After refactoring, verify:
- Root file is minimal - Under 50 lines, only universal info
- Links work - All referenced files exist
- No contradictions - Instructions are consistent
- Actionable content - Every instruction is specific
- Complete coverage - No instructions were lost (unless flagged for deletion)
- Self-contained files - Each linked file stands alone
04 同仓库更多 Skills
wealthville-pool-scoring
Score DeFi liquidity pools before providing liquidity. Returns a 0-100 pool score and an Enter/Hold/Exit/Reduce/Avoid verdict with calibrated confidence for ~68,800 Solana pools and 575 EVM pools, plus a public miss-inclusive track record. Use when asked whether a pool is worth LPing into, which pools look best right now, or how accurate the scores have actually been.
technical-debt-auditor
Scans a codebase and produces a quantified technical debt report with a prioritized remediation plan. Analyzes dead code, TODO/FIXME/HACK markers, dependency staleness, complexity hotspots, code duplication signals, test coverage gaps, unused exports, and outdated patterns. Use this skill when the user asks to audit their codebase for tech debt, wants a health check of their code quality, says things like "where should I focus refactoring", "find code smells", "audit my codebase", "what technical debt do we have", "prioritize cleanup work", or asks about code quality metrics. Also trigger when the user mentions "code rot", "legacy cleanup", "refactoring priorities", or wants to understand the overall health of a project before a major effort. Trigger even if the user doesn't say "technical debt" explicitly — any request to systematically assess code quality and identify improvement priorities belongs here.
readonly-sql-guard
Make an agent's database access provably read-only. Use when an agent runs caller-supplied SQL against a production/ERP/analytics database (a "query" MCP tool, text-to-SQL, an admin console), or when reviewing DB-access code for write/DDL/side-effect/transaction-escape risks. Checks what a statement CALLS, not just its shape.
web-artifacts-builder
Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
discernment-nudge
After you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.