ship-it

Runs an autonomous build loop against machine-checkable success criteria until the feature passes live QA.

Updated Sep 10, 2026
One-click install
npx skills add https://github.com/joydai2026-del/skills --skill ship-it-joydai2026-del
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ship-it
Source: https://github.com/joydai2026-del/skills/tree/main/ship-it
Command: npx skills add https://github.com/joydai2026-del/skills --skill ship-it-joydai2026-del

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Autonomous builds often stop early, fake completion on green unit tests, or halt mid-run to ask for secrets and permissions. This Skill front-loads every human-only input, writes a verifiable gate first, then loops until the real user-facing surface actually works. ## Core Features & Use Cases - Pre-flight input collection: Gathers secrets, credentials, spend consent, and test fixtures in one batch before the loop starts, verifying secret presence without printing values. - Verifier-first build loop: Generates a qa-checklist.yaml of automated checks as the stop condition, then iterates with bounded retries and anti-gaming guards. - Live-surface QA loop: Drives the real URL, bot, or CLI with browser automation until every live check passes, not just one happy-path click. - Use Case: Tell the agent "/ship-it add Stripe webhook handling" and it scopes the work, collects your API keys up front, builds against a checklist, runs verification, and hands back a ready-to-test summary. ## Quick Start Ask the agent to run /ship-it followed by a one-line feature description, then answer the single batch of pre-flight questions and wait for the ready-for-you summary.

Frequently Asked Questions about ship-it

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an autonomous build loop with an AI agent?▼

Invoke /ship-it with a one-line feature description. The skill collects all human-only inputs in one pre-flight batch, writes a qa-checklist.yaml gate first, then loops build and verification until every check passes on the live surface.

How does the agent avoid stopping mid-build to ask for secrets?▼

A Step 0.5 pre-flight scans the plan and codebase to predict every needed credential, access grant, and consent, then asks for all of them in one batch. Secrets are placed in .env by the owner and only their presence is verified, never printed.

Can the build loop run forever or fake completion?▼

No. Every loop is bounded: a check failing the same way three times is marked blocked with a reason. The skill forbids weakening or deleting tests to force green, and done requires both a green gate and a passing live-QA loop on the real surface.

When should I not use an autonomous ship-it workflow?▼

Do not use it for deploying a finished PR, looping an existing qa-checklist, or planning an unscoped feature. Use your ship workflow for deploys and a planning pipeline first if the feature is not yet scoped.

Does ship-it handle features with LLM or AI components?▼

Yes. Step 0 flags any AI surface, and Step 4 runs an ai-done calibration with evals. The quality band and rollback approval remain owner sign-offs; the agent proposes them but does not invent them.