Shipwright Harness

Commands Reference

Every slash command shipped by the Shipwright plugin, grouped by category. The Plugin highlights the nine commands you’ll reach for most often — this page catalogs all of them, including the full test-readiness pipeline and the maintenance patrol commands that run on a schedule.

Each command is invoked as /shipwright:{name} inside a Claude Code session with the plugin installed. Where a command is a thin wrapper around a same-named skill, it’s cross-linked to that skill’s entry in Agent Skills.

Core / Delivery Loop

The commands that drive the spec → plan → execute → review → deploy loop.

CommandWhat it does
/shipwright:prdInteractive PRD session — asks qualifying questions, researches context, and produces a PRODUCT-SPEC.md ready for /plan-session. (/shipwright:brainstorm is a deprecated alias that forwards to this command.)
/shipwright:plan-sessionEngineer planning pass — reads the product spec, explores the codebase, flags complexity, and produces a task queue in the task store. (/shipwright:plan-sesh is an alias.)
/shipwright:dev-taskExecute a specific task from the task store by id — build the feature, simplify, verify requirements, and ship a PR.
/shipwright:reviewReview a specific open PR — deep single-pass review with inline comments, policy-controlled posting.
/shipwright:patchAddress unresolved review findings, merge conflicts, and failing CI on a specific own open PR — queries GitHub directly, fixes in a worktree, pushes.
/shipwright:deployMerge and deploy a PR through the Deploy → Canary → Promote pipeline.
/shipwright:metricsAnalyze pipeline metrics — fix cascade trends, quality rates, and actionable recommendations across planning sessions.
/shipwright:researchLoad relevant project docs and web research for a given task, via an isolated sub-agent.
/shipwright:research-docsAnalyze the codebase and generate or update project documentation — like /init but for the docs/ directory. Supports an unattended --auto mode for cron use.

Test Readiness Pipeline

A five-phase pipeline that audits and improves test coverage end to end. Run in sequence — each phase reads the previous phase’s output.

CommandWhat it does
/shipwright:test-inventoryPhase 1 — inventory the codebase and prescribe the test layer (unit/integration/smoke/E2E) for every meaningful unit of code, ranked by criticality. Wraps the test-inventory skill.
/shipwright:test-designPhase 2 — design the ideal test system greenfield (frameworks, local substitutes, canary contract, speed budgets, CI shape) given the Phase 1 inventory. Wraps the test-design skill.
/shipwright:test-migrationPhase 3 — reconcile existing tests against the Phase 1 inventory and Phase 2 blueprint, bucketing each into reuse / promote / rebuild / net-new. Wraps the test-migration skill.
/shipwright:test-roadmapPhase 4 — synthesize the three prior artifacts into a single executable roadmap (test-readiness-plan.md) with sequenced milestones and an agent-executable task list. Wraps the test-roadmap skill.
/shipwright:test-fixPhase 5 — read test-readiness-plan.md and queue its flat T-NNN task list as task-store tasks (dependency edges included), one per row. Requires /test-roadmap to have been run first.
/shipwright:test-debtPost-execution analysis — compute the corrective-commit ratio per milestone from git log and flag under-specified milestones as planning debt for the next roadmap run. Wraps the test-debt skill.

Maintenance Patrols

Scheduled or on-demand scans that find and fix code entropy, production errors, and code duplication. Each pair follows the same shape: a -scan command that reports only, and a -fix command that reads the scan’s report and queues task-store tasks.

CommandWhat it does
/shipwright:entropy-scanScan the codebase against your golden principles and generate entropy-report.md. Report only — no code changes. Wraps the entropy-scan skill.
/shipwright:entropy-fixRead entropy-report.md and queue pr_worthy violations as task-store tasks, one task per golden principle, with per-finding HITL classification. Requires /entropy-scan first. Wraps the entropy-fix skill.
/shipwright:error-scanQuery the Sentry Issues API for unresolved issues, diff against the error-patrol ledger to find what’s new or regressed, and generate error-report.md. Report only — no code changes. Wraps the error-scan skill.
/shipwright:error-fixRead error-report.md, fetch each New/Regressed issue’s Sentry detail and stack trace, classify HITL per issue, and queue task-store tasks (plus companion observability-fix tasks where instrumentation gaps hinder root-causing). Requires /error-scan first. Wraps the error-fix skill.
/shipwright:error-resolveRead the error-patrol ledger, check each linked task’s current status via the task store, and resolve the corresponding Sentry issue once its gating task has reached deployed/done. Wraps the error-resolve skill.
/shipwright:consolidation-scanDiscover emerging duplicate/similar code patterns via judgment-driven comparison and track them across runs. Report only — no code changes. Wraps the consolidation-scan skill.
/shipwright:consolidation-fixRead consolidation-report.md and queue ready_to_propose duplication patterns as task-store tasks, one task per pattern, with a strangler-fig execution plan and per-finding HITL classification. Requires /consolidation-scan first. Wraps the consolidation-fix skill.

Planning & Research

Interactive and batch commands for discovery, spec authoring, doc maintenance, and cross-session learning capture.

CommandWhat it does
/shipwright:refresh-planRefresh stale planning doc tasks against the current codebase — verify file paths, update dependencies, and regenerate context.
/shipwright:learnCapture a durable learning and write it to its permanent home — CLAUDE.md, the relevant skill, or the right place per plugin routing rules. No staging file, no separate promote step.
/shipwright:learn-dreamExtract cross-session learnings from past session transcripts in batch — the “dreaming” pattern for autonomous agents with no human to react to a correction in real time.

Other

CommandWhat it does
/shipwright:hitlExecute a human-in-the-loop task — loads task context from the task store, assists with hands-on infra execution (terraform, helm, kubectl, gcloud — no tool restrictions), and marks the task done on exit.

Next steps

Continue to Agent Skills for the autonomous behaviors these commands wrap, or back to The Plugin for the plugin’s architecture and anatomy.