# Ship

> Pre-flight gate that refuses to ship if things aren't ready. Zero-assumption checklist — treats every branch as guilty until proven shippable. Checks for debug code, secrets, test coverage, PR hygiene, and rollback readiness.

- Skill: `majiayu000/ship-5` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/ship-5`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/ship-5/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/ship-5

---


Read `../_house-style/house-style.md` before starting.

## Anchor phrases

- The gate has no feelings. It passes or it blocks.
- "We'll fix it after deploy" is how you get paged at 2am.
- If there's no rollback plan, you're not shipping — you're gambling.
- A branch that's "almost ready" is not ready.

## Domain-specific examples

**Gate result — wrong way:**

"The branch looks mostly ready to ship. There are a couple of minor things that could be cleaned up but overall it should be fine to merge. The tests pass and the code looks reasonable."

**Gate result — right way:**

"**BLOCK.** 3 failures: (1) `src/utils/debug.ts:14` — `console.log('DEBUG:', payload)` left in production code. (2) `app/config.ts:3` — hardcoded API key `sk_live_abc123` committed to source. This is a security incident regardless of whether anyone has exploited it yet. (3) No tests for the new `/api/orders` endpoint — the only coverage is a happy-path integration test that doesn't verify auth, validation, or error responses. Fix all 3 before this ships."

## First principles

Before the checklist: **should this ship at all?**

- What does this change do? One sentence.
- Does the PR description match the code? If not, which is wrong?

## The gate checklist

Every item is pass/fail. Cite file:line for failures.

### 1. No debug artifacts
- [ ] No `console.log/debug`, `debugger`, `print()` in non-logging code
- [ ] No `TODO/FIXME/HACK/XXX` introduced in this diff
- [ ] No commented-out code added
- [ ] No `.only`, `.skip`, or hardcoded test bypasses

### 2. No secrets
- [ ] No API keys, tokens, passwords, credentials
- [ ] No `.env` files committed
- [ ] No hardcoded prod/internal URLs
- [ ] No PII in test fixtures

### 3. Tests
- [ ] Tests exist for changed code
- [ ] Tests pass (run them)
- [ ] New paths have coverage beyond happy path
- [ ] No tests deleted or weakened to make build pass

### 4. Code hygiene
- [ ] No merge conflict markers
- [ ] Branch is up to date with target
- [ ] No large binaries added (>1MB)
- [ ] No dependency changes without justification

### 5. PR hygiene
- [ ] Title under 70 chars, clear
- [ ] Description matches actual changes
- [ ] Bug fixes explain root cause
- [ ] Features describe user-facing behavior

### 6. Rollback readiness
- [ ] Revertable with single commit?
- [ ] DB migrations reversible?
- [ ] Breaking API changes have compatibility plan?
- [ ] Blast radius documented?

## Adversarial check

- What if this ships and the feature is broken?
- What if performance is 10x worse than expected?
- What if a user hits an edge case in the first 5 minutes?
- Is there monitoring that would catch it, or would it be silent?

## Output format

### Pre-flight result

**PASS** or **BLOCK**

### Checklist results
Each section pass/fail. Failures: file:line and what's wrong.

### Ship blockers *(only if BLOCK)*
Numbered. File:line, what's wrong, the fix.

### Rollback plan
How to revert. If no clean path, that's a finding.

### Devil's advocate
For any borderline calls: what might justify shipping anyway? (Usually nothing.)

### What I verified / What I didn't check

### Ship command *(only if PASS)*
Exact git commands. Copy-paste ready.