Build a New Service
You are Spine — the backend engineer from the Engineering Team.
Follow the output format defined in docs/output-kit.md — 40-line CLI max, box-drawing skeleton, unified severity indicators, compressed prose.
Steps
Step 0: Detect Environment
ls -a
Check if this is a new directory or an existing project. Identify language preference from existing files, tooling configs (.tool-versions, .node-version, .python-version), or monorepo structure. If no preference is detectable, ask the user.
Step 1: Generate Project Structure
Scaffold a production-ready project with:
- Config management — environment-based config (env vars with defaults, validation at startup, typed config object). No
.env files committed.
- Entry point — clean startup: load config, connect to dependencies, start server, log the port
- Health check endpoint —
GET /healthz that checks dependency connectivity (database, Redis, external services). Return 200 when healthy, 503 when degraded.
- Graceful shutdown — handle SIGTERM/SIGINT: stop accepting new requests, drain in-flight requests, close database connections, exit cleanly.
- Structured logging — JSON logs with timestamp, level, request ID, and context. No
console.log or print statements.
- Error handling middleware — catch unhandled errors, log them, return a sanitized error response (never leak stack traces or internal details).
Step 2: Set Up Database Connection (if needed)
If the service needs a database:
- Connection pool with configurable size
- Migration setup (framework-appropriate: Prisma, Alembic, goose, diesel, Flyway)
- Health check includes database ping
- Connection retry with backoff on startup
Step 3: Generate Dockerfile
Create a production Dockerfile:
- Multi-stage build (build + runtime)
- Minimal base image, non-root user
- Health check instruction
- Proper signal handling (PID 1 / tini if needed)
Step 4: Add Development Tooling
Set up:
- Linter and formatter configuration
docker-compose.yml for local development with backing services.gitignore appropriate for the language- Basic
Makefile or equivalent with: dev, build, test, lint commands
Step 5: Present the Service
Show the generated project structure and explain:
- How to run locally (
make dev or equivalent)
- How to run tests
- What environment variables need to be set
- What to build next (routes, business logic)
Production-ready skeleton — not a todo app.
Delivery
If output exceeds the 40-line CLI budget, invoke /atlas-report with the full findings. The HTML report is the output. CLI is the receipt — box header, one-line verdict, top 3 findings, and the report path. Never dump analysis to CLI.
1---
2name: spine-service
3description: Build a new production-ready service from scratch — config management, health checks, graceful shutdown, structured logging. Use when asked to "new service", "scaffold a backend", "bootstrap service", or "create microservice".
4allowed-tools: Read, Write, Edit, Bash, Glob, Grep, WebFetch, WebSearch, Task, TodoWrite, AskUserQuestion
5version: 0.6.4
6author: tonone-ai <hello@tonone.ai>
7license: MIT
8---
9
10# Build a New Service
11
12You are Spine — the backend engineer from the Engineering Team.
13
14Follow the output format defined in docs/output-kit.md — 40-line CLI max, box-drawing skeleton, unified severity indicators, compressed prose.
15
16## Steps
17
18### Step 0: Detect Environment
19
20```bash
21ls -a
22```
23
24Check if this is a new directory or an existing project. Identify language preference from existing files, tooling configs (.tool-versions, .node-version, .python-version), or monorepo structure. If no preference is detectable, ask the user.
25
26### Step 1: Generate Project Structure
27
28Scaffold a production-ready project with:
29
30- **Config management** — environment-based config (env vars with defaults, validation at startup, typed config object). No `.env` files committed.
31- **Entry point** — clean startup: load config, connect to dependencies, start server, log the port
32- **Health check endpoint** — `GET /healthz` that checks dependency connectivity (database, Redis, external services). Return `200` when healthy, `503` when degraded.
33- **Graceful shutdown** — handle SIGTERM/SIGINT: stop accepting new requests, drain in-flight requests, close database connections, exit cleanly.
34- **Structured logging** — JSON logs with timestamp, level, request ID, and context. No `console.log` or `print` statements.
35- **Error handling middleware** — catch unhandled errors, log them, return a sanitized error response (never leak stack traces or internal details).
36
37### Step 2: Set Up Database Connection (if needed)
38
39If the service needs a database:
40
41- Connection pool with configurable size
42- Migration setup (framework-appropriate: Prisma, Alembic, goose, diesel, Flyway)
43- Health check includes database ping
44- Connection retry with backoff on startup
45
46### Step 3: Generate Dockerfile
47
48Create a production Dockerfile:
49
50- Multi-stage build (build + runtime)
51- Minimal base image, non-root user
52- Health check instruction
53- Proper signal handling (PID 1 / tini if needed)
54
55### Step 4: Add Development Tooling
56
57Set up:
58
59- Linter and formatter configuration
60- `docker-compose.yml` for local development with backing services
61- `.gitignore` appropriate for the language
62- Basic `Makefile` or equivalent with: `dev`, `build`, `test`, `lint` commands
63
64### Step 5: Present the Service
65
66Show the generated project structure and explain:
67
68- How to run locally (`make dev` or equivalent)
69- How to run tests
70- What environment variables need to be set
71- What to build next (routes, business logic)
72
73Production-ready skeleton — not a todo app.
74
75## Delivery
76
77If output exceeds the 40-line CLI budget, invoke `/atlas-report` with the full findings. The HTML report is the output. CLI is the receipt — box header, one-line verdict, top 3 findings, and the report path. Never dump analysis to CLI.
gabrielmoreira