From ai-coding-config
Automates code shipping: merges base branch, runs detected tests, performs multi-review with auto-fixes, commits, pushes, creates PR. Use /ship or say 'ship it'.
npx claudepluginhub technickai/ai-coding-config --plugin ai-coding-configThis skill uses the workspace's default tool permissions.
Complete shipping workflow. When the user says "ship it", DO IT — don't ask for
Automates end-to-end branch shipping: detects base, merges it, runs tests, audits coverage on changes, reviews diffs, bumps version, commits, pushes, creates PR.
Orchestrates shipping workflow: reviews code via candid-loop, runs install/build/tests, creates GitHub PR, optionally auto-merges and updates issues. Use to ship changes safely.
Automates feature branch release pipeline: pre-flight checks, merge main, run tests, pre-landing review, version bump, changelog, bisectable commits, push, and PR creation.
Share bugs, ideas, or general feedback.
Complete shipping workflow. When the user says "ship it", DO IT — don't ask for confirmation at every step. Only stop for: merge conflicts, test failures, ASK items from review, or errors.
/ship — Full workflow on current branch/ship --skip-tests — Skip test step (use when tests already passed)/ship --skip-review — Skip multi-review (use for trivial changes)/ship --quick — Skip both tests and review (docs-only, typo fixes)Branch check: If on main/master, STOP. Tell the user to create a feature branch first. Do not create branches for them — that's a decision about naming and scope.
Detect base branch: Check git remote show origin or common names (main, master,
develop). This is the merge target.
Check for uncommitted changes: If there are unstaged/staged changes, note what will be included. Stash them before merging if git would refuse the merge, then unstash after.
Use git rev-parse --abbrev-ref origin/HEAD (fast, local) to detect the base branch.
git fetch origin <base>
git merge origin/<base>
If merge conflicts occur, STOP and report them. Do not auto-resolve — merge conflicts require human judgment about which version is correct.
If already up-to-date, continue silently.
Detect the test runner from project context:
package.json scripts: test, test:unit, test:integrationpytest / python -m pytestgo test ./...cargo testmix testbundle exec rspecRun the detected test command. If tests fail:
If no test runner is detected, skip with a note.
Skip this step if --skip-tests or --quick was passed.
Invoke /multi-review balanced (or deep for changes touching 10+ files).
Multi-review runs parallel agents, automatically fixes unambiguous issues, and reports what was Fixed, Wontfix, and Deferred. Any auto-fixed changes will be included in the commit.
If review surfaces items requiring user decision (ASK tier), STOP and present them before continuing.
Skip this step if --skip-review or --quick was passed.
Only if the project has these files:
VERSION file: Detect change magnitude from diff:
CHANGELOG.md: Auto-generate entry from diff. Use the commit messages and PR description to write a human-readable summary. Prepend to the existing changelog.
If neither file exists, skip silently.
Analyze all changes (original work + auto-fixes + version bump).
Create commits following the project's commit style (check recent git log for
conventions). If the repo uses conventional commits, follow that. If it uses emoji
prefixes, follow that.
For large changesets, create bisectable commits — logical units, not one giant commit. Typical split: migrations → models → services → controllers → tests → version/changelog.
For small changesets, one commit is fine.
Always include:
Co-Authored-By: Claude <noreply@anthropic.com>
git push -u origin <current-branch>
If push fails due to divergence, report it. Do not force-push.
First check if a PR already exists: gh pr view --json url. If it exists, skip creation
and output the existing PR URL. Otherwise create with --base <base>:
gh pr create --title "<concise title>" --base <base> --body "$(cat <<'EOF'
## Summary
<1-3 bullet points describing what this PR does>
## Changes
<Brief description of key changes, organized logically>
## Review findings
<Summary from multi-review: auto-fixed items, user decisions, wontfix items>
<Omit this section if review was skipped>
## Test plan
- [ ] Tests pass locally
- [ ] <Specific test scenarios relevant to the changes>
🤖 Generated with [Claude Code](https://claude.com/claude-code)
EOF
)"
Output the PR URL and a one-line summary:
PR #123 'Add user notifications' created on feature/notifications → main
https://github.com/org/repo/pull/123
This skill is designed to be invoked and left alone. The user said "ship it" — they want it shipped, not a conversation about shipping. Do the work, report the result.
The only interruptions should be genuine blockers that require human judgment, not confirmations of things the user already decided by saying "ship."