yonatangross/orchestkit
10 skills6k installs1.8k starsGitHub
Install
npx skills add https://github.com/yonatangross/orchestkitSkills in this repo
1Architecture Decision RecordArchitecture Decision Record is a reusable agent skill template for documenting significant technical choices in the OrchestKit style. Solo and indie builders use it when a fork in the road—database, auth model, monolith vs services, hosting—needs a durable record that future-you (and agents) can read without replaying Slack threads. The skill is not invocable as a one-shot command; it is procedural knowledge the agent applies when you ask for an ADR or when planning needs explicit rationale. Sections walk from context and conflicting forces through a crisp “we will” decision, implementation strategy, and balanced consequences. That pattern supports validate-time scoping, build-time design commits, and operate-time changes when you supersede an old ADR. Pair it with brainstorming or writing-plans when the decision follows a structured options review rather than a snap judgment.715installs2Devops DeploymentDevOps & Deployment is an OrchestKit workflow skill that gives solo builders structured guidance for CI/CD, containers, orchestration, and infrastructure as code. It activates when you are wiring GitHub Actions, writing Dockerfiles, targeting Kubernetes with Helm, or codifying cloud resources in Terraform—and it orients around real repo paths like workflows, k8s manifests, and tfvars. The skill pairs with a ci-cd-engineer agent persona and expects Claude Code 2.1.76+ compatibility, with an optional Docker presence check before container work. It covers GitOps-flavored flows and release strategy thinking, not just a single happy-path deploy script. Use it during ship when you promote artifacts to staging or production, and during operate when you iterate pipelines or infra. It is guidance-heavy with read/search tooling rather than a turnkey hosted platform—your agent still edits files in your repository.703installs3Ui ComponentsUI Components (Radix Accessibility Audit Checklist) gives solo builders a repeatable WCAG-oriented review pass for Radix UI primitives. Instead of guessing whether dialogs trap focus or menus announce state, you walk categorized checklists covering keyboard navigation, focus management, ARIA attributes Radix usually provides, screen reader behavior, and dialog-specific requirements like accessible names via Dialog.Title. Use it while you are still shaping frontend in Build, and again in Ship when accessibility is a launch blocker. It does not generate components or replace automated axe scans—it standardizes human-and-agent review so indie SaaS and extension UIs meet common compliance expectations. Pair it with your component PRs so regressions in overlays, menus, and form errors get caught before users do.690installs4Responsive Patternsresponsive-patterns is a compact reference skill for solo builders implementing modern responsive UI without relying solely on viewport media queries. It documents CSS container queries—how to establish containment on a parent, write `@container` conditions (including named containers and logical dimensions), and use container-relative units so typography and layout scale inside cards, dashboards, and split views. The readme also shows Tailwind CSS v4’s native container-query classes, including unnamed and named containers, so you can colocate responsive behavior with utility-first workflows. Use it while building SaaS dashboards, marketing blocks inside CMS layouts, or browser extensions with variable panel widths. It does not replace a full design system; it gives copy-paste patterns and vocabulary so your agent codes consistent, container-aware CSS instead of guessing breakpoint behavior.683installs5Rag RetrievalRAG Retrieval is a quality-assurance agent skill for solo builders shipping agentic products that depend on semantic search. It walks through a practical checklist spanning embedding configuration, chunk size and overlap, metadata filters, and top-k tuning, then extends into document grading prompts, query rewriting and decomposition, HyDE for vocabulary mismatch, and corrective web fallback when corpus retrieval fails. Self-RAG patterns cover when to retrieve at all, reflection tokens, and confidence thresholds so simple queries do not pay retrieval latency. Generation sections enforce context formatting, mandatory citations, hallucination detection, and sensible output length, while error handling expects graceful degradation and explicit fallback copy. Use it while implementing a new RAG stack in Orchestkit or similar frameworks, and again before ship when you need a pass/fail audit of retrieval precision, false positives, and production failure modes—not as a replacement for your embedding vendor docs, but as an agent-readable gate before you trust user-facing answers.604installs6Zustand PatternsZustand Patterns is a reference agent skill for OrchestKit that documents how to manage React application state with Zustand 5.x—the slice composition, Immer updates, shallow selectors, persist middleware, async actions, and devtools hooks indie SaaS teams typically need without adopting Redux. Solo builders install it when an agent is touching `store/` directories and you want consistent TypeScript-first patterns instead of one-off global singletons. The skill is model-invocation gated and frontend-agent scoped, so it behaves like embedded procedural knowledge rather than a user-facing chat command. It emphasizes seven documented core patterns plus explicit anti-patterns, which helps agents avoid prop-drilling hacks and unstable selector subscriptions during build. Use it while scaffolding dashboards, multi-step wizards, or extension popups where persisted UI state and derived selectors matter, then keep the reference handy through ship-phase refactors.552installs7Domain Driven DesignDomain-Driven Design from OrchestKit is a structured checklist skill for solo builders and small teams designing backends that match the business language instead of leaking persistence jargon everywhere. Use it when you are naming bounded contexts, drawing context maps, or implementing entities, value objects, aggregates, and repositories in Python-style stacks with PostgreSQL 18 conventions such as `gen_random_uuid_v7()`. The skill spans strategic design (ownership, integration patterns, glossary-aligned ubiquitous language) and tactical design (non-anemic entities, frozen value objects, aggregate sizing, repository interfaces). It fits Prism as a quality gate during backend build and again during ship review when architecture drift shows up in PRs. It is a checker-style artifact, not a code generator—your agent applies each box to the system you are already building.542installs8Agent OrchestrationAgent-orchestration is a framework selection checklist for solo builders who need to choose among LangGraph, CrewAI, OpenAI Agents SDK, Microsoft Agent Framework, and AG2 without betting the product on hype. It walks you through defining the use case, assessing single- versus multi-agent complexity, and documenting state management, human-in-the-loop, and observability requirements before you compare frameworks. Each option gets explicit when-to-pick criteria—complex stateful graphs, role-based crews, OpenAI-native handoffs and guardrails, Azure enterprise compliance, or community AutoGen flexibility—plus technical gates on expertise, documentation, and production readiness. The skill does not implement orchestration; it forces a written decision record so your Claude Code or Cursor sessions align on one stack. Use it early when architecture is still fluid and you want a repeatable evaluation ritual instead of copying the latest demo repo.537installs9MemoryMemory is an OrchestKit agent skill for unified read-side operations against your knowledge graph and project memory: search past decisions, load session context, sync-related flows, history, and Mermaid visualization via subcommands. It is built for solo builders who run Claude Code with the memory MCP server and want cross-session continuity without mixing in write paths—that belongs to /ork:remember. Triggers emphasize recall language (what did we decide, search memory, prior decisions) and explicitly avoid implement and remember anti-triggers. Paths cover .claude/memory and per-project memory directories. Use it during Build when grounding an agent in prior architecture choices, during Ship review when tracing why a decision was made, and during Operate when revisiting monitoring or iteration context from earlier sessions.534installs10Security PatternsSecurity Patterns packages a practical authentication security checklist for solo founders who own their own auth stack and cannot afford a missed cookie flag or stale OAuth grant. It walks password hashing choices (Argon2id preferred), brute-force limits, access and refresh token lifetimes, rotation on password change, and session cookie attributes that belong in every SaaS ship checklist. OAuth 2.1 expectations—PKCE for all clients, banned legacy grants, CSRF state, HTTPS-only endpoints—are spelled out so agent-assisted refactors do not reintroduce implicit flows. Optional passkey and MFA sections cover WebAuthn replay protection and TOTP backup codes. Use it while implementing Build backend auth, then again in Ship before release and during Operate when you change identity providers. Agents apply it as a checker-style pass over your design or PR description rather than generating secrets or running scans.463installs