Fifteen skills that make Claude's answers production-ready — from architecture decisions to deploy-day security — plus a four-agent auditor that reads your codebase in parallel and tells you exactly what's missing.
Shipwise knows which phase you're in and loads only what that phase needs. You never pick a skill — the work you're doing picks it.
Validate the idea, scope the MVP, form the entity, plan for accessibility before a line of UI exists.
Architecture decisions, fullstack patterns, CI/CD, test strategy, and security hardening as you write.
Observability, SEO, billing, legal compliance, and a launch-day sequence from T-2h to T+4h.
Event taxonomy, funnels, retention cohorts, revenue analytics, lifecycle email, cost control.
Fourteen domain skills plus an orchestrator that routes between them.
Four parallel auditors, a delta scanner, a fallback, and a gap analyser.
Every one carries at least one copy-paste template. Change the marked values and ship.
Session context, edit whispers, deploy gate, progress tracking. Zero friction.
Across three developer archetypes and three test fixtures. All open source.
/shipwiseNine questions, then a codebase scan and a project profile card with your readiness score, top gaps, and time estimates. The scan starts in the background after question four, so it's done before you are.
hooksFire on session start, file edits, and deploys. Auth code draws security tips, billing code draws webhook reminders, secrets draw a gitignore warning — deduplicated across restarts, once per topic. The deploy gate stops you shipping with P0 gaps open.
15 domainsYou don't invoke these. Claude loads them when the task matches — auth code pulls security patterns, payment code pulls billing practice. Off-topic prompts load nothing at all.
/launch-auditFour agents scan your codebase in parallel and merge into one readiness state. quick mode reads only what changed since the last run. /launch-checklist drills into one domain.
/launch-audit fans out to four specialised agents that read
different domains at the same time, then merges their findings into a single readiness state.
All four run on Haiku with a turn cap — which is the only reason a fan-out this wide is cheap
enough to run often.
/launch-audit
All four auditors at once across every domain. If one fails or times out, that domain falls back to the monolithic auditor — a partial failure never yields a partial audit.
/launch-audit quick
A delta auditor reads git diff and scans only changed files, returning just the items that improved or regressed. Past fifty changed files it upgrades itself to a full scan.
/shipwise
During first-run setup the audit starts in the background after question four, finishing while you're still answering the interview. You never wait on it.
Every reference doc carries at least one copy-paste template. Counts below are reference documents per skill, excluding the skill definition itself.
One prompt — "set up user auth with email/password, include proper security practices" — run against the same fixture project with and without the plugin. A 106% improvement, covering all eleven security dimensions the checklist tracks.
Full results, prompts and findings are open source.
| Beginner | Intermediate | Senior | |
|---|---|---|---|
| Checklist | Full explanations | Standard terms | Terse, code-first |
| Whispers | Jargon explained | One-line reminders | Suppressed if obvious |
| Deploy gate | Guided remediation | Time estimates | Gap list only |
| Audit | One recommendation | Options + tradeoffs | Decision matrix |
Set during scaffold, changed any time with /shipwise set-level.
Focus on shipping. Skip SOC 2, skip multi-region — they aren't your problem yet.
Add error tracking, basic monitoring, and a privacy policy.
Rate limiting, load testing, incident response, backup and disaster recovery.
Full observability, SOC 2 readiness, multi-region, cost optimisation.
Expected user scale reweights every priority in your checklist.
From the Claude Code plugin marketplace.
Nine questions and a codebase scan. Once per project.
Hooks are already live and skills auto-trigger. Audit before each milestone.
25/25 test scenarios · 11/11 security dimensions · zero friction
MIT · Built for Claude Code · Skills portable via convert.sh