freev0.1.0unverified
Ship Team Starter — 3 native subagents + 1 command, no conversion scripts
A curated software-shipping team as native Claude Code subagents: copy two
folders, type /ship, get architect → implementer → reviewer with a real gate.
Three roles, one command, zero install tooling.
Quick start
Claude Code: copy agents/ into .claude/agents/ and skills/ship/ into
.claude/skills/. Then: /ship add dark mode to settings.
What you get
| File | Role |
|---|---|
agents/architect.md | designs the smallest plan that ships; outputs a task list, never code |
agents/implementer.md | builds exactly the task list; small diffs, runs checks |
agents/reviewer.md | adversarial gate; can send work back, can't be skipped |
skills/ship/SKILL.md | /ship <feature> — runs the pipeline with explicit handoffs |
Why this instead of 232 free personas
The excellent agency-agents
(127k stars) gives you 16 divisions of personas — but in a generic format that
needs conversion scripts, with playbooks you chain by hand, and no benchmarks.
This pack is the opposite trade: **3 roles instead of 232, native format, wired
orchestration, and a proof ledger.** Curation is the product.
The paid bundle (coming) adds security review, QA, and delivery-tracking roles —
the full 7-role pipeline — plus per-model-release re-verification.
Proof
Unbenchmarked — v0. First benchmark: same feature via /ship vs single-agent
one-shot, measuring review-caught defects and rework diff size.
Attribution
Role design informed by agency-agents by AgentLand Contributors (MIT). All
prompt text in this pack is original.
Changelog
- 0.1.0 (2026-07-04) — initial release, unverified.
One email when a pack ships or gets re-benchmarked on a new model. No noise.