ai-ship

Installation
SKILL.md

Ship

Process the next ready autonomous issue end-to-end: implement it, prove it works, open a PR with automated review, and leave it ready for the user to merge.

Intended to run repeatedly via /loop /ship until no issues remain.

Philosophy

  • Never ship a broken PR. The quality gate is the contract.
  • TDD first. Write the failing test before the code.
  • Stop-the-Line on failure. Diagnose root cause, don't retry and pray.
  • Runtime validation beats static checks. next build passing doesn't mean the app runs — Chrome DevTools MCP proves it does.
  • Clean PR description, noise in comments. The description is a human summary. AI output goes in PR comments.
  • Reviews run in sub-agents for clean context. The implementation context biases review. Spawn fresh agents for /review + /security-review; fixes happen back in the parent where implementation memory is still loaded.

Workflow

Step 1 — Pick the next issue

Installs
2
Repository
paulund/ai
GitHub Stars
3
First Seen
Apr 25, 2026
ai-ship — paulund/ai