megastruktur

Ship It Skill — Productivity skill for Claude Code

Productivity community

Cross-platform coding-agent skill (Claude Code / OpenCode / omp): disciplined phased workflow with mandatory ask-tool STOP gates, adversarial plan critique, parallel execution planning, and an escapes.

How to install Ship It Skill

This entry records only its repository, not the path inside it, so there is no exact command to give. Open megastruktur/ship-it-skill and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Ship It Skill does

Alternatives in Productivity

Octoboss Reorganize Todos — You are the Octoboss — a cross-tentacle orchestrator 1.4k ★
Repurpose — Workflow recipe — turn one blog post, video, or idea into a full platform-native content pack (thread, LinkedI 1.3k ★
Cezar — Cezar is a parallel coding agents orchestrator 174 ★

README

ship-it — Disciplined Phased Workflow Skill

A cross-platform coding-agent skill that ships one non-trivial unit of implementation work end-to-end, with mandatory human decision gates. Runs on **Claude Code**, **OpenCode**, and **omp (oh-my-pi)**.

What it does

Every "I built the wrong thing" and "it merged and broke main" outcome comes from an agent skipping a confirmation step. ship-it makes those steps non-skippable:

Phase 0:    Read spec + investigate
Phase 1:    STOP — restate understanding, await confirmation      ← MANDATORY
Phase 2:    STOP — clarifying questions, await answers            ← if any
Phase 3:    STOP — plan saved to .ship-it/plans/, await approval  ← MANDATORY
Phase 3.5:  STOP — adversarial critic review (optional quorum)    ← MANDATORY (opt-out only)
Phase 4:    Implement (branch + todos + parallel subagents)
Phase 5:    Verify (lint / format / types / tests / E2E gates)
Phase 6:    Commit (only after gates pass)
Phase 7:    Roborev review loop (if installed)
Phase 8:    STOP — user verification + manual test guide          ← MANDATORY
Phase 9:    Update docs
Phase 10:   Merge per recorded strategy (squash local / PR)
Phase 11:   Report completion + end-user changelog

Key mechanisms:

STOP gates via the ask tool — decisions (go / fix, address / accept / quorum, merge / changes / hold) are delivered as structured questions with clickable options, not free text the user has to format correctly. Fallback to end-of-turn for agents without an ask tool.
Plan as artifact — plans are saved to .ship-it/plans/ in a canonical format, then audited by an independent critic subagent (self-preference bias is the enemy; "comfort is the tell"). Optional 3-domain quorum (architecture / security / executor) for infra or security-sensitive work.
Parallel Execution Plan — every plan must analyze which tasks can run as parallel subagent tranches; the critic auto-rejects plans that skip it.