Every workflow the console uses is a plain agent skill installed globally for your coding agent. Invoke them directly from Claude Code, Cursor, or opencode — no console required.
On your first shipper run (or via shipper skills), skills are installed into ~/.claude/skills/, ~/.cursor/skills/, or ~/.config/opencode/skills/ depending on your agent.
Explores your codebase, asks clarifying questions, and writes a phased markdown plan to .shipper/plans/open/. Also supports module URLs — install a Shipper module and plan building it into your repo.
e.g. “/shipper-plan https://shipper.is/modules/customer-support”
Orchestrates an entire open plan in one chat. Spins up a fresh subagent per phase that runs shipper-build, monitors progress, and continues until the plan moves to .shipper/plans/done/. This is what the web console Build button uses.
e.g. “/shipper-loop on .shipper/plans/open/my-feature.md”
Implements a single phase of an open plan in one agent session — checks off tasks and writes Completion Notes. Use when you want one phase at a time; prefer shipper-loop to finish the whole plan.
e.g. “use shipper-build on .shipper/plans/open/my-feature.md Phase 2”
Small one-off feature: plan and build in a single agent session.
e.g. “use shipper-spike to add a copy button to the hero”
Scaffolds a reviewable pull request from a completed plan — what changed, how to verify, and known risks. Creates the PR via gh.
e.g. “use shipper-ship on .shipper/plans/done/my-feature.md”
Evidence-first bug catalog and fix workflow. Reproduce before diagnosing, then drive the fix to proof in .shipper/bugs/.
e.g. “use shipper-bug to fix the login redirect loop”
Plans are committed markdown in .shipper/, so console users and direct-skill users work from the same files.