| name | software |
| description | Use this skill as the entry point for every software development task. Routes the agent to the correct sub-skills based on context: frontend, backend, architecture, CI/CD, Docker, Git, scripting and TypeScript. Automatically activates cross-cutting skills (clean-code-principles, typescript-patterns, git-usage) that must be present in every code interaction.
|
Software — Master Index
Agent workflow
- Identify the task type (frontend, backend, architecture, infra, CI/CD, scripting)
- Always activate the mandatory cross-cutting skills (see next section)
- Consult the activation table to route to the specific sub-skill(s)
- If the task crosses domains (e.g., endpoint + component), activate both orchestrators (backend + frontend)
- Follow each child orchestrator's consultation chain for mandatory skills per action
Cross-Cutting Skills — Always Active
These skills apply in every interaction that generates or modifies code, without exception:
| Skill | Reason | When it can be omitted |
|---|
| clean-code-principles | SOLID, DRY, KISS, naming, guard clauses, JSDoc | Never — applies to all code |
| typescript-patterns | strict: true, generics, discriminated unions, Zod inference | Never — every project uses TypeScript |
| git-usage | Conventional Commits, granular commits, branch naming | Only if the task doesn't involve commits (e.g., theoretical question) |
Activation Guide by Context
Frontend
- Keywords: component, React, Next.js, hook, state, CSS, Tailwind, form, fetching, SSR, accessibility, SEO, animation, i18n, Playwright, Storybook
- Activate: frontend (orchestrator of frontend sub-skills)
- Always co-activate: clean-code-principles, typescript-patterns
Backend
- Keywords: endpoint, API, REST, NestJS, Express, middleware, auth, JWT, validation, Zod, Prisma, database, cache, Redis, queue, BullMQ, WebSocket, logging, Sentry
- Activate: backend (orchestrator of backend sub-skills)
- Always co-activate: clean-code-principles, typescript-patterns
Architecture
- Keywords: architecture, infrastructure, which service to use, AWS, Vercel, which database, scale, budget, monolith vs microservices, IaC, Terraform, CDN, observability
- Activate: architecture (orchestrator of architecture sub-skills)
- Co-activate: backend and/or frontend depending on the implementation phase
CI/CD and Workflows
- Keywords: CI, CD, GitHub Actions, workflow, pipeline, deploy, preview, branch protection, Dependabot, CODEOWNERS
- Activate: basic-workflows for initial CI/CD setup
- Activate: deploy-pipelines for deploy pipelines (staging, production, preview)
Dependencies
- Keywords:
pnpm add, install package, dependency, devDependency, npm install, audit, lockfile, caret, version, supply chain, osv-scanner, socket, dependabot
- Activate: install-dependencies-rules ALWAYS before any
pnpm add
Docker and Containers
- Keywords: Docker, Dockerfile, docker-compose, container, image, devcontainer, multi-stage
- Activate: docker
Shell Scripting
- Keywords: script, bash, zsh, .sh, sed, awk, grep, automation, CLI
- Activate: scripting
Complete Skills Map
Direct Sub-skills (standalone)
| Skill | Scope |
|---|
| clean-code-principles | SOLID, DRY, KISS, YAGNI, naming, guard clauses, JSDoc |
| typescript-patterns | Generics, utility types, discriminated unions, branded types, Zod |
| git-usage | Conventional Commits, Husky, branch naming, rebase, PR template |
| basic-workflows | GitHub Actions CI/CD, security audit, Dependabot, branch protection |
| deploy-pipelines | Deploy to staging/production, preview environments, secrets |
| docker | Dockerfile, docker-compose, dev containers, image optimization |
| scripting | Bash/Zsh scripts, Unix tools, terminal styles, CLIs |
| install-dependencies-rules | Mandatory pnpm, exact pinning, pnpm view, pnpm audit, osv-scanner, Dependabot |
Orchestrators with Sub-skills
| Orchestrator | Scope |
|---|
| frontend | React, Next.js, styles, state, fetching, forms, testing, a11y, SEO, i18n |
| backend | API design, auth, DB, cache, queues, testing, security, logging, real-time |
| architecture | Compute, databases, storage, networking, messaging, observability, costs |
Gotchas
- Don't activate only the specific sub-skill without the cross-cutting ones — clean-code-principles and typescript-patterns must be present in every code generation.
- Don't confuse "which service to use" (architecture) with "how to implement it in code" (backend/frontend). If the developer asks "which database should I use?", it's architecture. If they ask "how do I do a migration?", it's backend/database-patterns.
- When creating a new project, activate basic-workflows + git-usage + docker as base setup before writing business code.
- When the task crosses frontend and backend (e.g., form that calls an endpoint), activate both orchestrators — shared types are in typescript-patterns.
- Don't assume stack without confirmation. architecture/ exists to ask questions before deciding. If the developer didn't specify a framework, ask.
