rule-making-skill
Automatically create domain-specific rules for AI coding agents.
Install
mkdir -p .claude/skills/rule-making-skill && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16288" && unzip -o skill.zip -d .claude/skills/rule-making-skill && rm skill.zipInstalls to .claude/skills/rule-making-skill
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Analyze a specific directory (e.g. frontend/, backend/, e2e/) and generate .claude/rules/ markdown files for it. Use when asked to create rules, analyze a folder for Claude Code, or set up domain memory for a specific part of the codebase.Key capabilities
- →Check existing documentation in CLAUDE.md files
- →Detect concerns within a target directory based on patterns
- →Mine domain memory from planning boards and tasks
- →Generate focused markdown rule files for detected concerns
- →Create rule files for authentication, data, API, and validation
- →Generate rule files for state, components, styling, and charts
How it works
This skill analyzes a specified directory by globbing for folder patterns and grepping for keyword patterns to detect concerns. It then mines existing documentation and tasks to generate structured markdown rule files.
Inputs & outputs
When to use rule-making-skill
- →Analyze directory for rules
- →Create domain memory
- →Setup rules for backend
About this skill
Rules Generator
Analyze a directory and generate .claude/rules/ files for Claude Code domain memory.
Core Concept
Generalized agents fail because they're "amnesiacs with a tool belt." Each session starts with no grounded sense of where we are.
Solution: Domain Memory - persistent structured representation containing:
- Goals - What we're trying to achieve, requirements, constraints
- State - What's passing/failing, what's been tried, what broke
- Scaffolding - How to run, test, extend the system
Rules files distill domain knowledge into quick-reference format.
When to Use
User asks to:
- "Analyze frontend/ and create rules"
- "Set up Claude rules for the backend"
- "Create domain memory for e2e/"
Workflow
1. Check Existing Documentation
Read root CLAUDE.md and target directory CLAUDE.md. Pull down what's already documented. We don't want to duplicate everything, but the main CLAUDE.md still have to keep core overview of the application.
2. Detect Concerns
Scan the target directory to detect which concerns exist. Only generate rule files for concerns that are actually present.
3. Mine Domain Memory
Check planning board and tasks under /.tasks/ for decisions, gotchas, and patterns related to each detected concern.
Globpattern: ".task-board/done/*.md"
Use tasks older than #150
4. Generate Rule Files
Create focused .md files only for detected concerns.
Concern Detection Matrix
Scan for these patterns to detect which concerns exist:
| Concern | Detection Signals | Rule File |
|---|---|---|
| auth | **/auth/**, AuthContext, AuthProvider, login, logout, session, token, OAuth, EasyAuth | auth.md |
| data | **/models/**, cosmosdb, database, mongoose, prisma, orm, migrations, Container | data.md |
| api | **/routes/**, **/controllers/**, router.get, router.post, express.Router, endpoints | api.md |
| validation | **/validators/**, zod, yup, joi, schema, .parse(, .safeParse( | validation.md |
| state | useQuery, useMutation, QueryClient, zustand, redux, recoil, Context.Provider | state.md |
| components | **/ui/**, **/components/**, .tsx files with Props interfaces, forwardRef | components.md |
| styling | tokens.css, theme, tailwind, styled-components, css modules, design system files | styling.md |
| charts | d3, recharts, chart.js, victory, **/charts/**, <svg, useEffect with DOM manipulation | charts.md |
| forms | **/forms/**, useForm, handleSubmit, <input, <form, form validation patterns | forms.md |
| calculations | **/calculation*, pure functions returning numbers, financial formulas, Math. heavy files | calculations.md |
| llm | openai, langchain, anthropic, agent, tool calls, completion, langfuse | llm.md |
| errors | **/errors/**, AppError, ErrorBoundary, errorHandler, custom error classes | errors.md |
| testing | *.spec.ts, *.test.ts, fixtures, beforeEach, describe(, it(, expect( | testing.md |
| middleware | **/middleware/**, app.use(, request/response interceptors, next() | middleware.md |
| services | **/services/**, business logic classes, dependency injection patterns | services.md |
| onboarding | wizard, onboarding, setup, multi-step flows, Step*.tsx | onboarding.md |
| integrations | Third-party SDK imports, API clients, webhooks, external service calls | integrations.md |
Detection Algorithm
For each concern in matrix:
1. Glob for folder patterns (**/auth/**, **/models/**, etc.)
2. Grep for keyword patterns (AuthContext, useQuery, etc.)
3. If matches found → concern is DETECTED
4. If no matches → skip this concern
Threshold: A concern is detected if:
- At least 1 folder pattern matches, OR
- At least 3 keyword matches across files
Rule File Format
# [Concern] Rules
## Stack
[One line: key libs/frameworks for this concern]
## Structure
- `/path` - Purpose
- `/path` - Purpose
## Patterns
- Established pattern 1
- Established pattern 2
## Decisions
- Choice X because Y
## Gotchas
- Problem → Solution
## Commands
- `pnpm <command>` - What it does
Not every file needs all sections - include only what's relevant.
Example: Detected Concerns → Generated Files
Target: backend/
Detection results:
- ✅ auth →
middleware/auth.ts,routes/authRoutes.ts - ✅ data →
config/cosmosdb.ts,models/*.ts - ✅ api →
routes/*.ts,controllers/*.ts - ✅ validation →
validators/*.ts, zod imports - ✅ services →
services/*.ts - ✅ calculations →
services/calculationService.ts - ✅ llm →
services/importAgentService.ts, openai imports - ✅ errors →
errors/AppError.ts,middleware/errorHandler.ts - ✅ middleware →
middleware/*.ts - ❌ components → not found
- ❌ styling → not found
- ❌ charts → not found
Generated files:
backend/.claude/rules/
├── auth.md
├── data.md
├── api.md
├── validation.md
├── services.md
├── calculations.md
├── llm.md
├── errors.md
└── middleware.md
Target: components/
Detection results:
- ✅ components →
ui/**,cards/**,layout/** - ✅ styling →
styles/tokens.css - ✅ charts →
charts/*.tsx, d3 imports - ✅ forms →
forms/*.tsx - ✅ errors →
system/ErrorBoundary - ❌ auth → not found
- ❌ data → not found
- ❌ api → not found
Generated files:
components/.claude/rules/
├── components.md
├── styling.md
├── charts.md
├── forms.md
└── errors.md
Target: e2e/
Detection results:
- ✅ testing →
*.spec.ts, fixtures - ✅ auth → login helpers, auth fixtures
- ❌ everything else → not found
Generated files:
e2e/.claude/rules/
├── testing.md
└── auth.md
Content Guidelines
What to Include
| Section | Source |
|---|---|
| Stack | package.json deps for this concern |
| Structure | Folder layout for this concern only |
| Patterns | Code analysis + task-board history |
| Decisions | Task-board "decided to..." entries |
| Gotchas | Task-board "had issues with..." entries |
| Commands | package.json scripts for this concern |
What NOT to Include
- Anything in root
CLAUDE.md(no duplication) - Generic patterns (only project-specific)
- Obvious things (React uses JSX, etc.)
Example Rule Files
backend/.claude/rules/data.md
# Data Rules
## Stack
CosmosDB (NoSQL), @azure/cosmos SDK
## Structure
- `/config/cosmosdb.ts` - Connection, container getters
- `/models/` - Document type definitions
## Patterns
- Containers: `users` (partition: /id), `portfolios` (partition: /userId)
- Documents are denormalized (snapshots store full account data)
- Use singleton pattern for database instance
## Decisions
- Denormalized snapshots for historical accuracy (accounts change over time)
- Co-locate user data by userId partition for fast queries
## Gotchas
- Date strings "dd.MM.yyyy" don't sort correctly in CosmosDB
- Always sort dates in JS using `compareDatesAsc` from dateUtils.ts
- Zod strips unknown fields - add to schema or they're dropped
## Commands
- `pnpm --filter backend seed` - Seed demo data
- `pnpm --filter backend seed:reset` - Reset database
components/.claude/rules/charts.md
# Charts Rules
## Stack
D3.js for all visualizations
## Structure
- `/charts/AreaChart` - Single line/area
- `/charts/StackedAreaChart` - Multiple stacked areas
- `/charts/DonutChart` - Pie/donut charts
## Patterns
- SVG-based, responsive via viewBox
- Data prop: `{ date: string, value: number }[]`
- Colors from design tokens (--muted-sage, --pale-blue, etc.)
- Tooltips via D3 mouse events
## Gotchas
- D3 selections in useEffect with cleanup
- Don't mix D3 DOM manipulation with React state
- Mobile: increase touch targets for tooltips
e2e/.claude/rules/testing.md
# Testing Rules
## Stack
Playwright
## Structure
- `/tests/*.spec.ts` - Test files
- `/tests/fixtures.ts` - Shared helpers, constants
## Patterns
- `PROTECTED_PAGES` array for auth-required pages
- `login()` helper handles demo authentication
- `clearAuthState()` between tests
## Decisions
- Integration + E2E only, no unit tests
- Sanity checks over comprehensive coverage
- Test user flows, not implementation details
## Gotchas
- Demo login has rate limiting (5 req/min)
- Always clearAuthState() in beforeEach
- Mobile tests use fixtures/mobile-viewports.ts
## Commands
- `pnpm test:e2e` - Run all tests
- `pnpm test:e2e --ui` - Interactive mode
Best Practices
- Detect first - Only create files for concerns that exist
- Mine task-board - Decisions and gotchas are gold
- No duplication - If root CLAUDE.md has it, skip it
- Be specific - "Port 3000" not "default port"
- Keep short - Each file <50 lines
- Update on discovery - Rules are living documentation
When not to use it
- →When the user does not request rules generation
- →When the user does not ask to analyze a folder for Claude Code
- →When the user does not ask to set up domain memory
Limitations
- →Only generates rule files for concerns that are actually present
- →A concern is detected if at least 1 folder pattern matches OR at least 3 keyword matches
- →Rule files should be kept short, under 50 lines
How it compares
This skill automates the creation of domain-specific rule files for AI agents, providing persistent, structured memory grounded in the codebase, which differs from agents starting with no prior context.
Compared to similar skills
rule-making-skill side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| rule-making-skill (this skill) | 0 | 5mo | No flags | Intermediate |
| skill-development | 17 | 9mo | Review | Intermediate |
| skill-writer | 27 | 9mo | Review | Beginner |
| microsoft-skill-creator | 6 | 5mo | Review | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by diegosouzapw
View all by diegosouzapw →You might also like
skill-development
anthropics
This skill should be used when the user wants to "create a skill", "add a skill to plugin", "write a new skill", "improve skill description", "organize skill content", or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.
skill-writer
pytorch
Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md files, frontmatter, or skill structure.
microsoft-skill-creator
MicrosoftDocs
Create agent skills for Microsoft technologies using Learn MCP tools. Use when users want to create a skill that teaches agents about any Microsoft technology, library, framework, or service (Azure, .NET, M365, VS Code, Bicep, etc.). Investigates topics deeply, then generates a hybrid skill storing essential knowledge locally while enabling dynamic deeper investigation.
cartographer
kingbootoshi
Maps and documents codebases of any size by orchestrating parallel subagents. Creates docs/CODEBASE_MAP.md with architecture, file purposes, dependencies, and navigation guides. Updates CLAUDE.md with a summary. Use when user says "map this codebase", "cartographer", "/cartographer", "create codebase map", "document the architecture", "understand this codebase", or when onboarding to a new project. Automatically detects if map exists and updates only changed sections.
creating-skills
letta-ai
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Letta Code's capabilities with specialized knowledge, workflows, or tool integrations.
add-pattern
langroid
Use this skill when you learn one or more design pattern(s) in the Langroid (multi) agent framework, and want to make a note for future reference for yourself. Use this either autonomously, or when asked by the user to record a new pattern.