AGENT SKILL

monorepo-coordination

Coordinate changes spanning 2+ repos of a multi-repo workspace or 2+ apps/packages of a monorepo -- merge-order plans, MR/PR templates, post-merge checklists. NOT for single-repo changes, even large ones.

License 许可
Apache-2.0
最近更新
2026-11-11
Stars
★ 8
Forks
5

01 前言 Front Matter

docs
{ "status": "reviewed", "updated": 1785974400000, "source_sha": "d9709fababf4" }
name
monorepo-coordination
owner
swarmery-core
version
1.0.0
description
Coordinate changes spanning 2+ repos of a multi-repo workspace or 2+ apps/packages of a monorepo -- merge-order plans, MR/PR templates, post-merge checklists. NOT for single-repo changes, even large ones.

02 资源文件 2

03 SKILL.md 正文

Purpose

Plan and sequence changes that span multiple repositories (or multiple monorepo apps/packages) so they land in the correct order, with CI-probe-enforced operator gates and post-merge validation of the whole system. Done when every repo is phased, each MR/PR carries dependency arrows and failure modes, and every operator step has a probe.

Rules

  • Read the repo shape from ${CLAUDE_PROJECT_DIR}/.claude/project.json (repos, monorepo) before planning — the phase model is identical for both shapes; only merge mechanics differ.
  • Merge order follows the phase model: 1 Foundation → 2 Operator action → 3 Wire consumers → 4 Consume. Never wire before foundation.
  • Every operator step gets a CI probe that fails if skipped — MR prose alone is never a gate.
  • When a contracted cross-service boundary changes, read the living contract document FIRST (see api-contract) and place its update no later than the emitter change.
  • Each MR is individually revert-safe; its description carries Depends on, Blocks, Operator steps, and Failure mode if merged out of order.
  • Use durable identifiers (file paths), never transient MR numbers; immutable digests, never mutable tags, in cross-repo handoffs.

Resources

  • Read resources/phase-model-and-procedure.md when building the plan — repo-shape detection, when (not) to use, inputs/outputs, the output template, and the six-step procedure with MR and CI-probe boilerplate.
  • Read resources/checks-and-examples.md before returning — self-check, common mistakes, escalation triggers, worked scenarios, failure modes, related skills.

How to use

What it does

Plans the landing order for a change touching multiple repositories — or multiple independently deployed monorepo apps: a phased merge order, MR/PR descriptions spelling out dependencies, CI probes that fail when a manual step was skipped, and a post-merge checklist.

When to use it

  • A change spans two or more repos, or monorepo packages that deploy independently.
  • A runtime env var seeded in one place, wired in another, read by the app.
  • A message-format change crosses a contracted device↔app boundary.
  • Not for single-repo changes (refactor-plan), hotfixes, dependency bumps, or image promotion.

How to invoke

Skill(skill: "core:monorepo-coordination")

Provide the change description and affected repos/packages (required); operator steps and existing MR links optional. Output is a plan capped at ~200 lines — nothing is merged or pushed.

Worked example

Change: add MAPS_API_KEY as a runtime env var. Repos: infrastructure,
deploy charts, main app. Operator step: run the secret bootstrap script.

The plan: seed the secret in the infrastructure repo, run the bootstrap script, wire the value into the chart with a placeholder guard, then read it in the app. Phase 2 gets a CI probe failing while the cluster secret contains CHANGE_ME; validation closes with a browser check that the value reaches the client.

04 同仓库更多 Skills