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.
npx skills add magic3007/dotfiles --skill herdr-coordinatorgit clone --depth 1 https://github.com/magic3007/dotfilesWrote 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.
[](https://agentmods.dev/skills/magic3007/dotfiles/herdr-coordinator)<a href="https://agentmods.dev/skills/magic3007/dotfiles/herdr-coordinator"><img src="https://agentmods.dev/badge/skills/magic3007/dotfiles/herdr-coordinator/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.
<a href="https://agentmods.dev/skills/magic3007/dotfiles/herdr-coordinator"><img src="https://agentmods.dev/badge/skills/magic3007/dotfiles/herdr-coordinator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00153 | $0.01909 |
| Opus 5.5 | $0.00061 | $0.00764 |
| Sonnet 5.5 | $0.00031 | $0.00382 |
| Haiku 4.5 | $0.00015 | $0.00191 |
Grade A, and why
herdr-coordinator 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 3d ago.
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.
How it starts
The opening of the file, as written. The whole thing — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
herdr-coordinator:协调者 + hc 后台 worker 工作流
本会话是协调者。你负责项目理解、规划、任务拆分、集成和终审。实现、调试、修测试、多文件改动交给后台 omp worker。
hc 的具体命令(启动、就绪判断、发送、等待、读回、核实模型、清理)全部以 herdr-desktop skill 为准,先读它。本 skill 只写流程和规则,不重复那些命令。
方案选择原则
每次选能同时改善用户体验(UX)、开发者体验(DX)和代理体验(AX)的方案,且不破坏现有功能。三者冲突时,先保证不破坏现有功能,再说明取舍。
worker 模型
模型不写死在命令里,取 ~/.omp/agent/config.yml 的 modelRoles 角色:
| 角色 | 命令 | 用途 |
|---|---|---|
worker |
omp --model worker |
实现、调试、修测试 |
reviewer |
omp --model reviewer |
独立审查、第二意见 |
- 只用 omp。不安装、不启动 Claude Code CLI 或 Codex CLI。
- 角色指向哪个 provider/model、档位是什么,由用户在 config 里定;启动前按
herdr-desktop第 0 步读出实际值并冒烟确认。角色缺失或模型不可用时停下告诉用户,不要换模型、换 provider 或降档位。
1. 先判断要不要委托
- 委托:大量实现、调试、修测试、多文件改动、需要独立审查。
- 自己做:单文件小改、只读问答、一两条命令能完成的事。为小事开 worker,启动和上下文交接的成本比活本身大。
2. 拆分并说明分工
先用一两段话告诉用户分工:几个 worker、各做什么、哪些并行、谁审查。然后在同一轮里把独立任务全部启动,不要启动一个等一个。
并发数按三项定,最多 10 个,不必用满:
uptime; nproc # load average 接近或超过 nproc 时少开
- 任务量:只有真正独立的切片才各开一个 worker。
- 配额:上一轮 worker 的 jsonl 里出现 429 或
errorMessage,就减少并发。
3. 隔离写入
- 并行改代码时,每个 worker 一个 worktree、一个分支。共享 checkout 中不允许两个 worker 写同一路径。
- worktree 从已提交的 HEAD 创建,协调者未提交的改动不会带过去。需要的话先提交,或在任务里写明。
- 本机没有全局 git 身份。委托前确认仓库有本地身份,worktree 会继承它:
git -C <repo> config user.email # 为空则 worker 无法提交,先告诉用户
git -C <repo> worktree add -b hc/<task> ../<repo名>-wt/<task> HEAD
worker 的 --cwd 就是这个 worktree 路径。
4. 任务合同
worker 看不到本会话,合同必须自足。每个任务都写全下面六项,并把第一条任务放进 hc send(就绪后发送):
目标:<要解决什么,为什么>
工作目录:<worktree 绝对路径>,分支 hc/<task>;只改 <路径范围>
上下文:<相关文件、接口约定、已知坑、运行/测试命令>
交付物:<在分支上提交的改动 + 简短说明>
完成标准:<可验证的条件,例如某测试通过、某命令输出>
回报:完成或卡住时执行 hc send <协调者地址> '<NONCE> <结果摘要 + 提交 sha>';长内容写文件,发路径
规则:
- 不要为没人要求的罕见情况或未来情况做设计。
- 不要改合同范围外的路径。
<协调者地址>写字面值(echo $HC_TERMINAL_ID的结果)。worker 自己的$HC_TERMINAL_ID指向它自己。- 每个 worker 用不同的 NONCE,后面靠它核对回报来源并找到会话 jsonl。
- 前端任务额外写:优化布局、视觉细节和实际交互,并在浏览器里自测。
- 不限制 worker 自己起 subagent 或 task。它需要并行探查、跑多个独立切片时,让它自己决定怎么分; 协调者只管合同范围和最终验收,多一层编排不需要额外授权。
5. 等待与反馈
- 用
hc wait --children all(或hc wait <name>)等待,不要 sleep 轮询。 - 返工意见用
hc send <原 worker>发回去,让它在原对话里接着改。重开 worker 会丢掉它已有的上下文。
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.
- 3d ago First seen · 128 lines · 153 tokens per session scan A 0c4c778047c7
herdr-coordinator is a skill published in the GitHub repository magic3007/dotfiles (11 stars, last pushed 2d ago), licensed MIT. It adds 153 tokens to every session and 1,909 once invoked, about $0.0006 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 2026-11-03.
herdr-coordinator compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| herdr-coordinatormagic3007/dotfiles | Skill | 11 | 2d ago | MIT | A | 153 | Claude Code, Codex |
| merge-readyurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 208 | Claude Code |
| sync-releaseurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 136 | Claude Code |
| prurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 161 | Claude Code |
| repo-statusurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 140 | Claude Code |
| setup-ciurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 119 | Claude Code |
| release-auditurmzd/dotfiles | Skill | 6 | 3d ago | Apache-2.0 | A | 126 | Claude Code |