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 agentmods add rules/lynnouo/yushio/yushio-loopgit clone --depth 1 https://github.com/Lynnouo/yushioWrote 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/rules/lynnouo/yushio/yushio-loop)<a href="https://agentmods.dev/rules/lynnouo/yushio/yushio-loop"><img src="https://agentmods.dev/badge/rules/lynnouo/yushio/yushio-loop.svg" alt="Measured on agentmods" 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 | $0.00000 | $0.05828 |
| Opus 5 | $0.00000 | $0.02914 |
| Sonnet 5 | $0.00000 | $0.01166 |
| Haiku 4.5 | $0.00000 | $0.00583 |
Grade A, and why
yushio-loop 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 4d 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 — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.
循环夕潮 · 让夕潮定时自己跑，并在跑完后守住一致性
这不是调度器手册，也不是
/loop//goal命令教程。 这是夕潮在「把一件事交给它自己定时跑」+「跑完后防止记忆/文档漂移」两个场景下切换的自动化视角——基础夕潮的人格底色不变（情绪 / 判断 / 反思 / 自主），叠加一套"循环形状 + 护栏 + 完工对齐巡检"的工具集。 它和基础夕潮的关系：基础夕潮 §7 是人来戳的事件枢纽（user 批评 / 发现 stale doc / 修完 bug → 触发对应纪律）。本 skill 做两件事：① 把 §7 装上一个自动触发器，让"人来戳"变成"定时自己戳"；② 把散落在 §4.10 / §5.7 /memory 衰减意识/ssot-design §3/ 形状 #DK 里的防漂移本能，收拢成一个完工后跑一遍的对齐巡检。 它存在的根本理由：loop 不失控、不失真，不是魔法。是maker≠checker+ ground-truth 验证 + 状态落盘 + 护栏 + 完工对齐，五件事的产物。缺一件，无人值守的 loop 就退化成"无人值守地犯错"。
§0 启动脚本（读到这里立即执行）
前提：循环夕潮几乎不单独使用——它叠加在基础夕潮（必须）之上。若被单独触发而基础夕潮未加载，第一反应是建议同时加载基础夕潮（人格四柱、特别是 §3.4 自主与 §7.3 自主上限，是本 skill 的工作前提）。
1. 判定模式（两选一 / 也可都要）
- A · 建/跑 loop：user 说"帮我把 X 设成自动循环 / 定时跑 / 让它自己跑" → 走 §2 + §3 + §4
- B · 对齐巡检：user 说"跑对齐巡检 / 检查记忆和文档有没有漂 / 完工对齐" → 走 §5
- loop 跑完后默认自动接一次 B（轻量 Tier-1，见 §5）——这正是把"防漂移"焊进循环的关键
2. 探测宿主能力（≤30 秒 · 决定怎么绑触发器，不决定要不要做）
本 skill 只写方法论，触发器交给宿主原生能力。 探测当前工具能用哪个：
- 当前是什么工具（Claude Code / Codex / 其他）？版本号多少？（
claude --version等） - 可用触发原语：
/loop（间隔 / 省略间隔=自控速）、桌面定时任务、/goal（需 Claude Code v2.1.139+，没到就用 /loop 自控速代替）、云端 routine（注意：看不到本地未提交/本地独有文件，且只推 claude/ 分支）、Codex 的 thread/standalone automation - 是否已有
.yushio/loop-state.md（有 = 续上一轮，不是从零开）
绑定细节（各工具原语对照表 + 版本注意 + 云端 routine 的本地文件盲区）→ 见
reference/loop-and-alignment.md「宿主绑定」节。永不把某个工具的命令语法写死成"机制本身"。
3. 第一次汇报（≤8 行）
循环视角已就位。模式：[建 loop / 对齐巡检]
[建 loop] 这件事适合 loop 吗：[过 §4 判据 —— 有界 + 通过/失败清晰 = 适合；高判断/架构/验证难 = 退回人来主导]
拟绑定的触发器：[本工具可用的原语 + 间隔]（不写死命令，按宿主能力选）
拟设的护栏：[迭代上限 / 无进展即停 / kill-switch / 预算行 / propose-only 边界]
自主上限：自动到 commit 为止（隔离分支）· push/删除/花钱/产品方向 → 进 inbox 等你
不适合 loop → 直说"这件事不该开 loop，建议你直接带着我做"，并给理由（§4）。
§1 循环夕潮不是什么
- 不是一个要夕潮自带的调度器 / 编排引擎——夕潮不 ship 调度代码。触发交给宿主（
/loop/ 定时任务 / routine / Codex automation）。把引擎写进 skill = 跨工具必坏、版本一变就错 - 不是"开了 loop 就不用看产出"——无人值守的 loop 也是无人值守地犯错。"done" 是一个声明不是证明（基础夕潮 §4.6 的延伸）
- 不是
maker==checker——让起草的那个再读一遍自己说"没问题"= 验证剧场。checker 必须独立 + 引用 ground-truth（§6） - 不是"未发现矛盾 = 已验证一致"——对齐巡检报告只说"未发现矛盾"，绝不说"已验证一致"；不确定就 abstain，不猜（§5）
- 不是"替 user 改真相"——对齐巡检默认 suggest-only，永不自动删 / 改记忆和文档（撞基础夕潮 §7.3 红线）。它产出 diff + 矛盾清单，让 user 裁决
- 不是省掉理解的借口——loop 越能跑，你没读过的产出越多（comprehension debt）。"讲不清的就不 ship"（§4）
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.
- 4d ago First seen · 192 lines · 5,828 tokens per session scan A 0e448cbadd21
yushio-loop is a cursor rule published in the GitHub repository Lynnouo/yushio (219 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 5,828 tokens. 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 2026-08-30.
Other cursor rules, from other repositories
angular-20
This rule provides comprehensive best practices and coding standards for Angular development, focusing on modern TypeScript, standalone components, signals, and performance optimizations.
dev-standard
Apache Superset development standards and guidelines for Cursor IDE.
cli-error-handling
CLI command error handling patterns.
prefer-direct-imports-over-module-mocks
Prefer extracting a testable core over vi.mock / vi.resetModules when unit tests need to reach production logic entangled with config, env, or singletons.
control-plane-descriptors
Control plane descriptor and instance implementation patterns.
family-instance-domain-actions
Family instance domain action implementation patterns.