qmu/ship

Ship

Merge a pull request, deploy to production, and verify the deployment. The agent acts as the deployment agent, following the ## Deploy and ## Verify instructions in the project's CLAUDE.md.

This skill is the trip-independent ship essence: it operates on the current branch's PR. Worktree handling and drive/trip context routing are not part of this skill — in Claude Code they are handled separately by the trip workflow and the /ship command. Any agent can run this skill directly to ship the current branch.

Agent Compatibility

This skill works on any Agent-Skills-compatible agent. Where a step uses AskUserQuestion (workspace/ticket guards, deploy confirmation), use the agent's native way of presenting a multiple-choice question (or ask in plain chat). The confirmations are mandatory; only the prompt mechanism varies. (This skill has no subagent fan-out.)

1. CLAUDE.md Convention

Deploy and verify instructions live in the project's CLAUDE.md — the file authored by the user that already carries the project's instructions. The ship workflow reads its ## Deploy and ## Verify sections to learn how to deploy and verify.

1-1. Search Order

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/find-claude-md.sh

Searches: ./CLAUDE.md

1-2. Expected Sections

## Deploy
Step-by-step deployment instructions for the agent to execute.

## Verify
Health checks, smoke tests, and expected outcomes.

1-3. Confirmation

Before executing deploy instructions, display the Deploy section and ask the user to confirm via AskUserQuestion. If the user declines, deployment is skipped.

1-4. Fallback

If no CLAUDE.md is found, or it has no ## Deploy section, skip deploy and verify steps. Inform the user that deployment was skipped because CLAUDE.md carried no deploy instructions.

2. Shell Scripts

2-1. Pre-check

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/pre-check.sh "<branch>"

Verifies a PR exists for the branch. Returns JSON with PR number, URL, and merge status.

2-2. Merge PR

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/merge-pr.sh "<pr-number>"

Merges the PR, checks out main, and pulls to sync. Returns JSON with merge status and commit hash.

2-3. Find CLAUDE.md

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/find-claude-md.sh

Searches for the project's CLAUDE.md. Returns JSON with path or {"found": false}.

2-4. Check Todo

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/check-todo.sh

Checks if .workaholic/tickets/todo/ has remaining tickets. Returns JSON with cleanliness status, count, and ticket list. Used as a pre-merge guard to prevent shipping with unfinished work.

2-5. Extract Carry-Overs

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/extract-carryover.sh "<branch>" "<pr-number>" "<pr-url>"

Reads the just-shipped story (.workaholic/stories/<branch>.md), parses each ### concern block in section 6 (Concerns), and writes one file per concern under .workaholic/concerns/ as <pr-number>-<slug>.md (with severity and a Title/Description/How-to-Fix body). Returns JSON:

{"status":"ok","extracted":10,"files":["..."]}

Commits the new files with message Carry over concerns from PR #<pr-number> so the corpus stays under version control. Skips silently when no story file exists or section 6 is empty.

3. Workspace Guard

bash ${CLAUDE_PLUGIN_ROOT}/skills/branching/scripts/check-workspace.sh

Parse the JSON output. If clean is true, proceed silently to the Ticket Guard.

If clean is false, display the summary to the user and ask via AskUserQuestion with selectable options:

• "Ignore and proceed" - Continue with the ship workflow. The unrelated changes will remain in the workspace after the command completes.
• "Stop" - Halt the workflow so you can handle the changes first.

If the user selects "Stop", end the workflow immediately.

4. Ticket Guard

bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/check-todo.sh

Parse the JSON output. If clean is true, proceed silently to the Ship Flow.

If clean is false, display the ticket list to the user: "Cannot ship: N ticket(s) remaining in .workaholic/tickets/todo/:" followed by the ticket filenames. Then ask via AskUserQuestion with selectable options:

• "Move all to icebox" - Move all remaining tickets to .workaholic/tickets/icebox/, stage and commit "Move remaining tickets to icebox", then proceed to the Ship Flow.
• "Stop" - Halt the workflow so you can handle tickets first (run /drive, manually reorganize, etc.)

If the user selects "Stop", end the workflow immediately.

5. Ship Flow

Ship the current branch's PR. (Worktree sync/cleanup and drive/trip routing are not here; in Claude Code those are handled by the trip workflow.)

1. Pre-check: Run bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/pre-check.sh "<branch>". If found is false: inform user "No PR found for this branch. Run /report first." and stop. If merged is true: skip to Deploy.
2. Merge PR: Run bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/merge-pr.sh "<pr-number>". On failure, inform user and stop.
3. Extract carry-overs: Run bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/extract-carryover.sh "<branch>" "<pr-number>" "<pr-url>". Persists active Concerns from the just-shipped story's section 6 into .workaholic/concerns/. Commits the new files. Skips silently when no story file exists or section 6 is empty. Report extracted count from the JSON output.
4. Deploy: Run bash ${CLAUDE_PLUGIN_ROOT}/skills/ship/scripts/find-claude-md.sh. If found is false, or CLAUDE.md has no ## Deploy section: inform user "No deploy instructions found in CLAUDE.md. Deployment skipped." and skip to summary. Otherwise: read CLAUDE.md, find the ## Deploy section, ask confirmation via AskUserQuestion, execute if confirmed.
5. Verify: If a ## Deploy section was found, read the ## Verify section of CLAUDE.md and execute. Report results.
6. Summarize: PR merge status (number, URL), carry-over extraction count, deployment status, verification results.