agentmods.dev

multi-agent-project-setup

multi-agent-project-setup is a skill for Claude Code, Codex from hxy91819/mason-skills. It costs 65 tokens per session (4,161 once invoked), scanned A, original, MIT.

A setup and review guide for projects where many coding agents work in parallel. It covers repository layout, checks before merging, continuous integration (automated build and test runs), releases, shared status, waiting, and monitoring.

In plain words
What is it for?
Use it to initialize or audit a multi-agent coding project, organize separate worktrees, divide checks between merge time and later, manage waiting processes, and track queues, delays, and machine health.
Why use it?
It helps prevent parallel work from blocking each other, rerunning unnecessary checks, overwriting shared files, or hiding stalled work. It also provides measurements and completion criteria for deciding whether the setup is working.

Skill for Claude CodeCodex

Written for Claude Code and Codex: disable-model-invocation in frontmatter, but also agents/openai.yaml present. Also seen: mentions AGENTS.md; $skill-name invocation.

not rated 3repo changed today A scan 65 tokens original MIT

Good fit Use it to initialize or audit a multi-agent coding project, organize separate worktrees, divide checks between merge time and later, manage waiting processes, and track queues, delays, and machine health.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/hxy91819/mason-skills/multi-agent-project-setup
View source ↗ hxy91819/mason-skills
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add hxy91819/mason-skills --skill multi-agent-project-setup
Clone the repo
git clone --depth 1 https://github.com/hxy91819/mason-skills

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for multi-agent-project-setup

README.md
[![agentmods](https://agentmods.dev/badge/skills/hxy91819/mason-skills/multi-agent-project-setup/github.svg)](https://agentmods.dev/skills/hxy91819/mason-skills/multi-agent-project-setup)
Your own site
<a href="https://agentmods.dev/skills/hxy91819/mason-skills/multi-agent-project-setup"><img src="https://agentmods.dev/badge/skills/hxy91819/mason-skills/multi-agent-project-setup/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for multi-agent-project-setup

Your own site · 80×15
<a href="https://agentmods.dev/skills/hxy91819/mason-skills/multi-agent-project-setup"><img src="https://agentmods.dev/badge/skills/hxy91819/mason-skills/multi-agent-project-setup.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 65 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,161 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00065 $0.04161
Opus 5.5 $0.00026 $0.01664
Sonnet 5.5 $0.00013 $0.00832
Haiku 4.5 $0.00006 $0.00416

Measured today against content hash 901d93308214, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

multi-agent-project-setup scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

common-skills/multi-agent-project-setup/SKILL.md · 124 lines

How it starts

The opening of the file, as written. The whole thing — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.

多 Agent 协作项目的初始化配置

单人项目的默认配置在十几个线程并行时变成一条单车道：每个 MR 跑全量 CI、master 一动就重跑、状态记在共享文件里、等待挂在会话里。一个四人团队带十几个 Agent 并行开发的平台项目 2026-11-05 实测：每天 130 个合入，工作 MR 从建到合入 p50 41 分钟、p90 106 分钟，一半 CI 容量花在记账 MR 上；数字与事故见实测证据，向用户出示依据或判断自己项目是否同病时再读。

每条配置过一遍二十线程测试：二十个线程同时这么做会怎样，会不会互相拖住。

原则

  • 门禁只保护别人：合入前的检查只为保证别的线程 rebase 到的主干能编译、单元测试绿；其余检查合入后跑，红了告警、向前修。没有生产用户时成立，开放用户后按环境重新评估。
  • 共享状态出仓库：会被多线程同时改的东西（工作簿、台账、证据）放本机被 git 忽略的目录，或拆成每阶段一个文件。
  • 等待不占会话：超过 20 分钟的等待放持久进程，结束时唤醒线程。
  • 规则进检查：能用 lint 或 CI 保证的约定做成检查，不写成提醒。
  • 环境只从主干来：共享环境的每类变更只有一条登记过的通道，排障只读。
  • 有数才有判断：队列、滞后、停摆、机器健康都有指标、阈值和告警去向。

流程

  1. 量基线，至少半天样本：合入数与构成（工作 / 发布 / 记账）、工作 MR 从建到合入 p50 与 p90、CI 墙钟与取消数、检查池并发与排队、发布滞后、同时活跃线程数、开发机 load。样本不足半天不下结论。
  2. 逐项对照清单，缺项按"做法"落地。整改顺序固定：先减少走全量 CI 的 MR 数，再减少每个 MR 的 CI 轮数，再把容器类检查挪到合入后，最后才加机器。
  3. 同口径复测，写进项目的 operations 记录。门禁改动先落地，再用自然流量验收，不为凑样本等。

完成判据：清单每行都有"已有 / 已落地 / 不适用（原因）"三选一，复测数字与基线并列。

配置清单

仓库与文档

做法 为什么 验证
主检出留主干，每个任务独立 worktree，交付后 git worktree remove 十几个线程改同一检出会互相覆盖 git worktree list 没有超过一天的闲置 worktree
.local/ 进 .gitignore，作本机状态目录（工作簿、证据、临时产物） 这些内容进仓库就要走 MR 和 CI git check-ignore .local/x 通过
AGENTS.md 只放常驻规则加"任务 → 先读"路由表，细节按任务加载；读文档先 rg 标题再按章节读 一次 cat 多份文档会截断，Agent 还要补读 AGENTS.md 一屏内，路由表每行一个任务
术语表加 ADR，名词只有一个写法 名字撞车让线程各改各的 新名词先进术语表再出现在代码
数据库迁移编号用 UTC 时间戳 顺序号在并行分支上抢号 迁移文件名匹配 YYYYMMDDHHMMSS_
会被多方追加的登记表和账本：一条记录一个文件，读取时组合；CI 禁止往汇总文件追加，加载器遇到重复键报错 并行 MR 都往文件末尾追加，反复冲突；有一次 git 合并没报冲突，YAML 里却多出同名顶层键，旧状态被静默覆盖 汇总文件只剩索引字段；两个分支互相 rebase，两个方向都无冲突
建 worktree 的工具从最新主干建分支，并打印落后的提交数和近 7 天规则文档的变化 一个线程落后主干 17 个提交，照旧规则跑了 18 分钟全量检查 新建 worktree 的输出里有落后数和规则变化

合入与 CI

做法 为什么 验证
合并结果 CI 一次绿灯即合入；master 漂移只评论留证，不重跑（双边都新增迁移时例外） 漂移重跑让每个 MR 跑两三轮，合入要一小时 合入工具日志里没有"重跑"
主干检查每次合入触发，正在跑的跑完，排队的只跑最新提交（superseded 跳过，不用 cancel-in-progress） 取消策略下合入密集时主干检查一次都跑不完 任一小时内主干检查至少一次成功
门禁三层：合入前只跑弹性公共池上的静态检查；改迁移、镜像、部署输入时才在合入前跑容器类；合入后主干检查跑全部 容器类检查占自有检查池，是排队的根源 纯 Go 或前端改动的 MR 墙钟 5–6 分钟
纯文档 MR 只跑链接检查，不占自有检查池；用于编译或测试输入的 .md 按例外清单走全量 文档 MR 走全量白占 14 分钟 文档 MR 墙钟 2 分钟内
统一合入工具：核对远端头、MR 快照、检查对应的 SHA 三者一致，快照陈旧用工具刷新 线程用 close/reopen 刷新快照，顶掉正在跑的构建 取消数里没有同 SHA 替换
一个阶段一个 MR，审查修复进同一 MR 一个任务开十几个 MR，每个都占一轮 CI 分支名对应的 MR 数为 1
自有检查池并发按机器 load 和内存定，排队中位超过 5 分钟再加节点 机器没满时加机器不解决问题 检查池 load、可用内存、排队中位三个数并列
主干红时，先按失败用例名或固定标题标记搜在途的修复 MR，有就等它合入再 rebase；自己修时标题带同一个标记 主干红时所有 MR 一起红，各自在自己的 MR 里修，会修出好几份 修主干的 MR 标题都带这个标记
按路径分流时，用 git diff --name-only -z 读取路径，只算 merge-base..HEAD；白名单之外的改动一律走全量 按行读取时中文文件名被转义，纯文档 MR 被判成代码改动；拿目标分支 HEAD 比较，把目标分支新增的代码也算了进来，白跑约 10 分钟检查 含中文文件名的纯文档 MR 走文档通道
MR 检查用 BASE_SHA 证明历史迁移没被改过，缺 BASE_SHA 就失败 历史迁移被改，库里已经执行过的版本和代码对不上 改历史迁移的 MR 必红
CI 平台声明的语义要实测：含 bash 语法的步骤显式用 bash -euo pipefail 调用；排队超时和执行超时分开设 公共池声明了 bash，实际仍用 /bin/sh 启动；构建机并发配小时，排队超过默认 10 分钟就报失败 ShellCheck 按实际解释器跑
测试前清掉 CI 事件变量，设 LANG/LC_ALL=C.UTF-8 和 TZ=UTC；测试用系统分配的端口 测试里新建的 git 仓库继承了 CI 变量，去远端 fetch，11 次主干检查红了 7 次；C locale 下中文报错不可读；固定端口互相冲突 满载重放 200 轮零失败
CI 机器上的临时容器和匿名卷经统一入口创建，打上"受管"和到期时间标签，每次作业开始先回收到期的；docker rm 必须带 -v 任务被取消或被 SIGKILL 时 trap 不执行；rm -f 不带 -v，9 天积了 1259 个匿名卷、274G 强杀后由下一次作业回收；检查脚本拦下不带 -v 的 rm

Read the full file on GitHub · 124 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. today Changed · +16 lines 901d93308214
  2. 2d ago First seen · 108 lines · 65 tokens per session scan A e92a8417e774

Subscribe to this mod's changes

multi-agent-project-setup is a skill published in the GitHub repository hxy91819/mason-skills (3 stars, last pushed today), licensed MIT. It adds 65 tokens to every session and 4,161 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-05.

CI and CDDevOps and cloudGit and workflow
Related

multi-agent-project-setup compared with skills from other repositories

multi-agent-project-setup next to 6 skills from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
multi-agent-project-setuphxy91819/mason-skillsSkill3todayMITA65Claude Code, Codex
deployment-patternsaffaan-m/ECCSkill299kyesterdayMITA41Claude Code, Kiro
agent-github-modesruvnet/rufloSkill81kyesterdayMITA21Claude Code
baby-sitlangchain-ai/open-sweSkill11ktodayMITA30Claude Code, Codex
chinese-git-workflowjnMetaCode/superpowers-zhSkill8.5k12d agoMITA69Claude Code
atmos-hookscloudposse/atmosSkill1.4ktodayApache-2.0A46Claude Code
atmos-procloudposse/atmosSkill1.4ktodayApache-2.0A38Claude Code