npx claudepluginhub agent-sh/ship --plugin shipCreate a versioned release with automatic ecosystem and tooling detection.
Use when monitoring CI and handling review comments during /ship. Details mandatory wait periods, auto-reviewer handling, and comment resolution.
Use when deploying and validating during /ship. Details Railway/Vercel deployment, smoke testing, and rollback procedures.
This file contains error handling procedures for `/ship`.
Complete PR workflow from commit to production with validation
End-to-end PR workflow - commit, push, create PR, monitor CI, address every review comment, merge, deploy, validate - plus discovery-first release automation.
Shipping code involves a long tail of manual steps after the code is written: staging, committing, pushing, writing PR descriptions, waiting for CI, reading reviewer comments, pushing fixes, re-waiting, merging, cleaning up branches. ship handles all of it. It also enforces a discipline most developers skip: addressing every single review comment before merge.
Use cases:
agentsys install ship
Requires agentsys runtime.
/ship # Full workflow: commit, PR, CI, review, merge
/ship --dry-run # Show what would happen without executing
/ship --strategy rebase # Use rebase merge instead of squash
/ship --base develop # Target a non-default branch
/release # Patch release (auto-discovers release method)
/release minor # Minor version bump
/release major --dry-run # Preview major release
/release --skip-publish # Tag and changelog only, no registry publish
/release --yes # Skip confirmation prompt
| Phase | Description |
|---|---|
| 1 | Pre-flight - detect CI platform, deploy platform, branch strategy, verify tools |
| 2 | Commit - stage changes (excluding secrets), generate semantic commit message |
| 3 | Create PR - push branch, open PR with summary and test plan |
| 4 | CI & Review Loop - wait for CI, wait 3 min for auto-reviewers, address all comments, iterate until zero unresolved threads (max 10 iterations) |
| 5 | Internal Review - 4 parallel review passes (standalone only; skipped when called from /next-task) |
| 6 | Merge - verify mergeable status, confirm zero unresolved threads, merge with chosen strategy |
| 7-10 | Deploy & Validate - platform-specific deployment, health checks, auto-rollback on failure |
| 11 | Cleanup - remove worktrees, close linked issues, delete branches |
| 12 | Report - final status summary |
Phase 4 always runs - even when invoked from /next-task. After PR creation, ship waits 3 minutes for auto-reviewers to post, then enters a loop:
The loop exits only when unresolved threads reach zero.
ship auto-detects your project's CI and deployment setup:
The release agent uses a discovery-first approach - it inspects your repo before executing anything.
Discovery order:
Supported ecosystems: npm, Cargo, Python (pyproject.toml, setup.py), Go, Maven, Gradle, RubyGems, NuGet, Dart, Composer, Hex, Swift.
Constraints: Tests must pass before tagging. Version bump is reverted if tests fail. Tags are never force-pushed.
| Agent | Model | Role |
|---|---|---|
| release-agent | Sonnet | Discover release method and execute versioned releases |
When called from the next-task workflow (via --state-file), ship skips its internal review (Phase 5) and deslop/docs steps since next-task already ran those. Phase 4 (CI & review loop) still runs because external auto-reviewers comment after PR creation.
gh) - required for PR operationsMIT
Team-oriented workflow plugin with role agents, 27 specialist agents, ECC-inspired commands, layered rules, and hooks skeleton.
Uses power tools
Uses Bash, Write, or Edit tools
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Comprehensive startup business analysis with market sizing (TAM/SAM/SOM), financial modeling, team planning, and strategic research
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification