Production-grade skills, rules, and prompts for AI coding agents (Cursor, Kiro, Claude Code, Windsurf, etc.)
A curated collection of agent skills, behavioral rules, prompting frameworks, and domain knowledge that transform generic AI coding assistants into disciplined, senior-level engineering partners.
This repository is organized into clear categories:
| Category | Purpose |
|---|---|
agent-rules/ |
Behavioral constraints and operating principles that govern how the agent works |
agent-skills/ |
Domain-specific knowledge packs that give the agent expertise in particular areas |
prompts/ |
Structured prompting frameworks and playbooks for task execution |
user-knowledge/ |
Reference documents and cheat sheets for the human user |
├── agent-rules/
│ ├── hard-rules.md # Core behavioral constraints (truth > agreement, no blind execution)
│ ├── post-change-audit.md # Mandatory safety audit after every code change
│ └── anti-sycophancy.md # No flattery, no filler — direct communication only
│
├── agent-skills/
│ ├── audio-app/ # Full skill pack for AI-powered audio web apps
│ │ ├── SKILLS-INDEX.md
│ │ ├── 00-operating-rules.md
│ │ ├── 10-audio-engine.md
│ │ ├── 20-waveform-and-stems.md
│ │ ├── 30-piano-roll-and-midi.md
│ │ ├── 40-conversion-workflows.md
│ │ ├── 50-beat-maker.md
│ │ ├── 60-ui-ux-system.md
│ │ ├── 70-package-validation.md
│ │ ├── 80-component-blueprints.md
│ │ ├── 90-implementation-guardrails.md
│ │ └── 99-agent-checklists.md
│ │
│ ├── pricing-monetization/ # SaaS pricing, paywalls, and monetization strategy
│ │ ├── README.md
│ │ ├── WORKFLOW.md
│ │ ├── MODULES.md
│ │ ├── TEMPLATES.md
│ │ ├── AGENT_EXECUTION_GUIDE.md
│ │ ├── ALTERNATIVES.md
│ │ ├── PRICING_STRATEGY_GUIDE.md
│ │ ├── IMPLEMENTATION_ARTIFACTS.md
│ │ └── modules/ # 14 individual module files (A-N)
│ │
│ ├── senior-engineering/ # Codebase audit & production-readiness skill pack
│ │ ├── SKILLS-INDEX.md
│ │ ├── 00-agent-rules.md
│ │ ├── 10-code-audit.md
│ │ ├── 20-completeness-check.md
│ │ ├── 30-quality-upgrade.md
│ │ ├── 40-ui-validation.md
│ │ ├── 50-auth-billing-security.md
│ │ ├── 60-security.md
│ │ ├── 70-env-docker-git.md
│ │ ├── 80-performance.md
│ │ ├── 90-testing-ci.md
│ │ ├── 95-production-readiness.md
│ │ └── 99-final-audit-checklist.md
│ │
│ ├── frontend-design/ # Premium frontend & UI/UX design skill
│ │ ├── frontend-design-skill.md
│ │ └── high-impact-ui-ux-strategies.md
│ │
│ └── software-craftsmanship/ # Clean architecture, TDD, DDD, SOLID
│ └── software-design-skill.md
│
├── prompts/
│ ├── 00-core-doctrine.md # The "constitution" — universal agent operating system
│ ├── 01-request-playbook.md # Standard task execution (build features, refactor)
│ ├── 02-refresh-playbook.md # Root cause analysis for persistent bugs
│ ├── 03-retro-playbook.md # Post-session self-improvement & doctrine evolution
│ └── 04-conciseness-directive.md # Radical conciseness overlay (stackable)
│
├── user-knowledge/
│ ├── communication-cheat-sheet.md # How to give clear, effective feedback to AI agents
│ └── huggingface-cli-reference.md # HF Hub CLI usage guide
│
├── README.md
└── LICENSE
- Pick what you need — Browse the categories above and grab the files relevant to your workflow.
- Load into your agent — Copy contents into your agent's rules, skills, or system prompt depending on your tool:
- Cursor:
.cursor/rules/or.cursor/skills/ - Kiro:
.kiro/steering/ - Claude Code:
CLAUDE.mdor project rules - Windsurf:
.windsurfrules - Other agents: System prompt or context files
- Cursor:
- Combine as needed — These are composable. Use rules + skills + prompts together.
| Goal | Load These |
|---|---|
| General vibe coding with guardrails | agent-rules/hard-rules.md + prompts/00-core-doctrine.md |
| Full-stack feature development | Above + agent-skills/senior-engineering/ |
| Premium UI/frontend work | agent-skills/frontend-design/ + agent-rules/post-change-audit.md |
| Audio/music web app | agent-skills/audio-app/ (full pack) |
| SaaS pricing & monetization | agent-skills/pricing-monetization/ (full pack) |
| Debugging persistent bugs | prompts/02-refresh-playbook.md |
| Codebase audit before launch | agent-skills/senior-engineering/SKILLS-INDEX.md (follow the workflow) |
These materials are built on five principles:
- Research first, always — Never act on assumption. Investigate before changing.
- Truth over agreement — The agent corrects you when you're wrong.
- Extreme ownership — Responsibility extends beyond the immediate task.
- Autonomous problem-solving — Exhaust all options before escalating.
- Premium quality by default — No "good enough." Ship what a senior engineer would approve.
Hard behavioral constraints. These tell the agent what it must and must not do regardless of what you ask. Think of these as the agent's "constitution" — non-negotiable operating principles.
Domain expertise packages. These give the agent deep knowledge in a specific area (audio engineering, frontend design, security auditing, etc.). Skills contain patterns, checklists, decision trees, and implementation guidance.
Structured task execution frameworks. These are templates you paste into chat to initiate specific workflows (building features, debugging, self-improvement). They enforce a rigorous sequence of phases.
Reference materials for you, the human. Cheat sheets, vocabulary guides, and tool references that help you communicate more effectively with AI agents.
This is Phase 1. Planned additions:
- Backend/API development skill pack
- Database & data modeling skills
- DevOps & infrastructure skills
- Mobile development skills
- Testing & QA skill pack
- More prompting playbooks (migration, refactoring, greenfield)
- Agent-specific setup guides (Cursor, Kiro, Claude Code)
- Example projects using these skills
Contributions welcome. If you have skills, rules, or prompts that make AI coding agents better, open a PR.
Guidelines:
- Follow the existing naming conventions and structure
- One concern per file — don't combine unrelated things
- Include practical examples, not just theory
- Test with at least one AI coding agent before submitting
CC BY 4.0 — Use freely, modify freely, but give credit.
Built for the vibe coding era. Make your agents better.