Restyle the description and when_to_use frontmatter of the whole skill catalog to the skills-engineering-autonomous pattern, so Claude Code and Codex route to the right skill: - description: capability + em-dash/colon differentiator (key invariant in CAPS) -> concrete sequenced mechanism -> explicit fail-closed/honesty clause -> a terse trailing "Use ..." (atomic skills) or a phase-chain/"Composes ..." closer (composites). - when_to_use: "Use when <positives + examples>. Do NOT use <negatives> -> redirect to the sibling skill by name; <crisp principle>." + explicit-user-only/confirm caveat where the skill mutates or spawns agents. Frontmatter-only: skill bodies are unchanged. Grounded each rewrite in the skill's actual behavior. (ship-playbook's restyle rode with its v1.16.0 commit.)
7.1 KiB
name, version, description, allowed-tools, argument-hint, arguments, when_to_use, effort
| name | version | description | allowed-tools | argument-hint | arguments | when_to_use | effort | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| nodejs | 1.1.0 | Write Node.js and Bun backend code the way THIS project already does it, not by generic defaults — a TypeScript-first reference that detects the runtime and carries the real conventions for HTTP servers, Zod boundary validation, typed error handling, pino logging, database access, auth, BullMQ queues, caching, async and graceful-shutdown patterns, security, CLI tooling, testing, and observability, so a change lands idiomatic and review-ready instead of merely running. Use when a task touches this project's Node.js or Bun backend and should follow its conventions rather than generic defaults. |
|
[Node.js/Bun backend task or subsystem] |
|
Use when the task touches Node.js or Bun backend code: HTTP servers, API endpoints, database queries, auth, jobs, caching, logging, testing, CLI tools, or deployment. Examples: "add an API endpoint", "fix this async error", "set up BullMQ jobs", "add request validation", "build a CLI command", "containerize this service". Do NOT use for frontend-only work, or when a framework owns the surface — Next.js (nextjs), NestJS (nestjs), or Laravel (laravel); this is the framework-agnostic backend surface. | high |
Non-negotiable rules:
- Read
references/stack.mdfirst to determine the runtime (Node.js or Bun), framework, and locked decisions. - Then load only the references needed for the actual task.
- TypeScript strict mode — no
any, no implicit returns, no unchecked index access. - All async functions must handle errors — no unhandled promise rejections. Use try/catch or
.catch()at every boundary. - Input validation at system boundaries — Zod schemas on every external input (request bodies, query params, env vars, CLI args). Never trust
req.body. - Structured logging via pino — no
console.login production code. JSON to stdout, parsed by log aggregators. - Graceful shutdown — handle SIGTERM/SIGINT, drain connections, finish in-flight work, close DB pools.
- No secrets in code or logs — env vars validated at startup, never logged, never in error responses.
nodejs
Inputs
$request: The backend task, subsystem, bug, or feature being worked on
Goal
Route Node.js/Bun backend work through the project's conventions so implementation follows the established patterns for server architecture, data access, error handling, and deployment.
Step 0: Read the stack contract
Always start with:
references/stack.md
That establishes: runtime (Node.js or Bun), framework (Express/Fastify/Hono/none), ORM, test runner, package manager, and locked dependency choices.
If bun.lockb or bunfig.toml exists, the runtime is Bun — also load references/bun-runtime.md for native API differences.
Success criteria: The project's runtime, framework, and toolchain choices are explicit before implementation starts.
Step 1: Load only the relevant references
Use the routing table to pick reference files. Do not bulk-load the full reference tree.
| Task | Read |
|---|---|
| Runtime, TypeScript, package manager, locked deps | references/stack.md |
| Folder conventions, entry points, monorepo layout | references/project-structure.md |
| Express/Fastify/Hono patterns, middleware, routing | references/http-server.md |
| REST conventions, versioning, pagination, error responses | references/api-design.md |
| Zod/AJV input validation, DTO patterns | references/validation.md |
| Prisma/Drizzle/Knex, migrations, connection pooling | references/database.md |
| JWT, sessions, OAuth2, RBAC, middleware guards | references/auth.md |
| Error classes, async error boundaries, HTTP error responses | references/error-handling.md |
| pino structured logging, request correlation, log levels | references/logging.md |
| vitest/jest/bun test, supertest, test factories, coverage | references/testing.md |
| Promises, streams, worker threads, AbortController, shutdown | references/async-patterns.md |
| helmet, CORS, rate limiting, input sanitization, dep audit | references/security.md |
| Redis, in-memory caching, cache invalidation patterns | references/caching.md |
| BullMQ, job patterns, retry strategies, dead-letter queues | references/queues-jobs.md |
| Env validation, dotenv, config modules, secrets management | references/config.md |
| OpenTelemetry, health checks, metrics, distributed tracing | references/observability.md |
| Multi-stage Dockerfile, .dockerignore, prod vs dev images | references/docker.md |
| ws/Socket.io, connection lifecycle, scaling, rooms | references/websockets.md |
| commander/yargs, argument parsing, exit codes, stdin/stdout | references/cli.md |
| Bun-native APIs, bun test, bun build, Bun.serve, Bun.$ | references/bun-runtime.md |
Multiple tasks? Read multiple files. The references are self-contained.
Success criteria: Only the task-relevant backend conventions are in play.
Step 2: Implement with the core backend guardrails
Keep these rules active:
- TypeScript strict mode with
noUncheckedIndexedAccess - all external input validated at the boundary (Zod schemas)
- errors are typed, caught, and returned as structured HTTP responses
- logging via pino child loggers with request correlation IDs
- database access through the project's ORM/query builder, not raw SQL strings
- mutations wrapped in transactions where atomicity matters
- graceful shutdown: SIGTERM handler drains server, closes pools, exits cleanly
- env vars validated at startup — fail fast on missing required config
- no
any, noastype assertions unless justified with a comment - if Bun runtime: prefer
Bun.serve(),Bun.file(),bun testover Node.js equivalents
Success criteria: The change matches the project's backend architecture instead of generic defaults.
Step 3: Verify the affected surface
Use the narrowest relevant verification:
- unit tests (
vitest run,jest, orbun test) - integration tests with supertest or actual HTTP calls
- type checking (
tsc --noEmit) - linting (
eslint .) - if Docker: build the image and verify it starts
Success criteria: The changed backend surface still builds, type-checks, and passes tests.
Guardrails
- Do not inline the whole Node.js handbook in
SKILL.md. - Do not skip
references/stack.md. - Do not use
console.login production code — use pino. - Do not bypass input validation at API boundaries.
- Do not leave unhandled promise rejections.
- Do not hardcode secrets, ports, or environment-specific values.
- Do not add
disable-model-invocation; this is a normal domain skill.
When To Load References
-
references/stack.mdAlways. -
references/bun-runtime.mdWhen the project uses Bun (detected viabun.lockborbunfig.toml). -
then only the task-relevant files under
references/
Output Contract
Report:
- which references were loaded
- the architecture pattern chosen
- the change made
- the verification run