Core Concepts

Skills

Reusable knowledge modules written in Codex `SKILL.md` format, grouped by category.

Skills live in .agents/skills/<name>/SKILL.md and should stay narrow, explicit, and reusable across repositories.

A skill can include references/, scripts/, and assets/, but it should never behave like hidden automation.

The shipped catalog is grouped below by Skill Categories so you can find the right skill quickly.

Skill Contract

Keep each skill narrow enough to be composable with other skills.
Use SKILL.md for the instruction contract.
Use references/ for deeper examples or templates.
Use scripts/ only for optional helpers that should be suggested, not silently executed.

Skill Categories

48 shipped skills organized by domain.

Planning & Routing

app-builder

New app scaffolding, stack selection, and high-level project setup.

architecture

Requirements, tradeoffs, ADRs, and system design decisions.

behavioral-modes

Explicit working modes such as brainstorm, implement, debug, or review.

brainstorming

Clarify scope and generate options before implementation.

intelligent-routing

Choose the best specialist skills or subagents for a task.

parallel-agents

Bounded delegation and parallel subagent coordination.

plan-writing

Written implementation plans, breakdowns, and checklists.

planning

Execution-ready planning with scope, risks, and acceptance criteria.

repo-onboarding

Fast map of an unfamiliar repository before making changes.

Backend & Platform

clean-code

Pragmatic coding standards and scoped implementation quality.

api-patterns

API design, response shapes, versioning, and protocol choices.

database-design

Schema design, migrations, indexes, and query strategy.

nodejs-best-practices

Node.js architecture, async patterns, and backend decision-making.

python-patterns

Python project structure, async choices, and framework direction.

rust-pro

Modern Rust systems work, async design, and performance.

mcp-builder

Design principles for building MCP servers, tools, and resources.

Frontend & UI

frontend-design

Web UI design systems, hierarchy, typography, and aesthetics.

mobile-design

Touch-first UX, mobile patterns, and platform conventions.

nextjs-react-expert

React or Next.js architecture, rendering, and performance.

tailwind-patterns

Tailwind CSS v4 patterns, tokens, and utility architecture.

web-design-guidelines

UI audits against structured web interface guidelines.

i18n-localization

Translations, locale files, RTL support, and hardcoded string checks.

game-development

Game-project routing and platform-specific game skill selection.

Debugging & Review

bug-hunt

Disciplined reproduction, scoping, and root-cause isolation.

debugging

Evidence-based debugging before changing code.

systematic-debugging

Structured 4-phase debugging with explicit hypotheses.

code-review

Patch and branch review for correctness and regressions.

code-review-checklist

Supplemental prompts and checks during code review.

high-signal-review

Findings-first review output focused on real risk.

Testing & Validation

lint-and-validate

Linting, type checks, formatting, and static validation.

tdd-workflow

RED-GREEN-REFACTOR test-driven development cycle.

test-hardening

Strengthen weak or flaky tests around critical behavior.

testing-patterns

Unit, integration, and mocking strategies.

webapp-testing

Browser testing, deep audits, and Playwright-style checks.

release-readiness

Higher-confidence validation for rollout and operational risk.

Docs, Delivery & Operations

doc

Work with `.docx` documents where formatting fidelity matters.

docs-shipper

Ship docs that match real product behavior and commands.

documentation-templates

README, API, and technical documentation structure guidance.

deployment-procedures

Safe deployment principles, verification, and rollback thinking.

server-management

Operational process management, monitoring, and scaling decisions.

mcp-onboarding

Evaluate, adopt, and roll out MCP servers safely.

Security, Performance & Discoverability

vulnerability-scanner

OWASP-aware vulnerability analysis and attack-surface review.

red-team-tactics

Authorized adversary-emulation and defensive reporting patterns.

performance-profiling

Measure-first profiling and performance optimization guidance.

seo-fundamentals

Search visibility, E-E-A-T, and Core Web Vitals basics.

geo-fundamentals

Optimization for AI search and citation engines.

Shell & Environment

bash-linux

Bash and Linux command patterns for macOS or Linux.

powershell-windows

PowerShell patterns, pitfalls, and Windows shell syntax.

Minimal Loading Rule

The default rule is simple: load as little as you can while still doing good work.

Choose the workflow first, then add only the skills that meaningfully improve the current task.

Skill Structure

SKILL.md: the instruction contract and main entrypoint for the skill.
references/: deeper examples, templates, or supporting guidance.
scripts/: optional helpers that should be invoked deliberately, not silently.
assets/: supporting files such as prompts, data, or media.
agents/openai.yaml: optional invocation policy for skills that bundle deeper routing behavior.
Some skills also include task-specific files such as verify.md, handoff.md, or checklists.