STAR: Command for Claude Code

.agents/commands/star-auto.zh-CN.md

star-auto.zh-CN is a command for Claude Code from wanghao9610/STAR. It costs 0 tokens per session (3,017 once invoked), scanned A, original, MIT.

A command that lets an automated research workflow continue through approved steps toward a stated goal, with an optional stopping rule and involvement level.

In plain words
What is it for?
Use it to run a research task across prepared skills and local code steps, optionally stopping when a condition such as a time limit or paid API call is reached.
Why use it?
It reduces repeated approval prompts during a planned workflow while recording actions and respecting limits such as cost, deletion, or system changes.

Command for Claude Code

Written for Claude Code: a Claude Code command (commands/*.md). Also seen: installed under .agents/ (shared by several agents).

This is wanghao9610/STAR's own configuration. It tells Claude Code how to work on STAR itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything STAR configures →

Reuse

Borrowing it

Nothing to install: this file belongs to wanghao9610/STAR. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/wanghao9610/STAR/main/.agents/commands/star-auto.zh-CN.md
Clone the repo
git clone --depth 1 https://github.com/wanghao9610/STAR

Made for: Claude Code.

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 star-auto.zh-CN

README.md
[![agentmods](https://agentmods.dev/badge/commands/wanghao9610/star/star-auto.zh-cn/github.svg)](https://agentmods.dev/commands/wanghao9610/star/star-auto.zh-cn)
Your own site
<a href="https://agentmods.dev/commands/wanghao9610/star/star-auto.zh-cn"><img src="https://agentmods.dev/badge/commands/wanghao9610/star/star-auto.zh-cn/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 star-auto.zh-CN

Your own site · 80×15
<a href="https://agentmods.dev/commands/wanghao9610/star/star-auto.zh-cn"><img src="https://agentmods.dev/badge/commands/wanghao9610/star/star-auto.zh-cn.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,017 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 $0.00000 $0.03017
Opus 5 $0.00000 $0.01509
Sonnet 5 $0.00000 $0.00603
Haiku 4.5 $0.00000 $0.00302

Measured 4d ago against content hash 0194b24357c1, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

star-auto.zh-CN 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.

.agents/commands/star-auto.zh-CN.md · 57 lines

How it starts

The opening of the file, as written. The whole thing — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.

朝目标自动推进工作流

本文件是 star-auto.md 的中文对照版。运行时以英文原文件为准。

用户敲下了 star-auto——这次调用运行的就是本文件。敲下它即规约 §10.7 所定义的自动授权:本次运行追逐目标期间,每个下一步动作点名的 skill 都可以直接启动,包括只许显式调用的那七个;备好的红线命令也可以直接启动而非交回。involve=low 时,同一份明确授权还覆盖下文列出的 executor 链本地实现、机械审查修复与 Git 生命周期动作;除此以外,被启动的运行仍与用户亲手点名时完全一致。

调用形状:star-auto <目标> [stop=<停止线>] [involve=<档位>]

解析调用

  • 先剥掉 involve=<档位>,只取停止线文本之外的——stop= 之前的,或整个调用末尾的那个 token;停止线里出现的同样字样是停止线文字,不是 token(规约 §7.7)。什么都没设档位时——没有 token、.env 里也没有 INVOLVE——本次运行解析为 low 而非 medium:自主运行对自己的裁量题取标注的推荐项并逐条记录(§7.8)。low 时,本次调用本身预先批准 executor 推荐的执行计划、按路径暂存与提交、创建执行分支或 worktree、不删除内容且不改变行为的审查修复、审查干净后的 squash 合并,以及仅在持久产物已迁出且 git status --porcelain 为空时移除 worktree。这些动作不发授权问题,全部留记录。需要删除内容的审查修复跳过并转交,不自行猜。合并后的执行分支默认保留且不发问,避免无人值守清理依赖强制删除。这项例外通过内部 token auto=unattended 传给 executor 与 code reviewer;它不属于用户调用格式。
  • 再剥 stop=:从这个 token 起到剩余文本末尾的全部内容就是停止线,用户自己的话写的自然语言——stop=any single training over 4 hours, or any paid API callstop=用卡超过 2 张就停(中英皆可,混写也行)。给了,它就是本次运行不得越过的线:启动一条备好的命令之前,先陈述其预期开销并逐条对照——越线的、或无法对照判断的,照 §2 原样打印交回。没给,本次运行不设上限:备好的红线命令自主启动,以完成目标为先(§2 的唯一例外)。无论哪种,每次启动都连同陈述的开销记入决策记录。除上面的干净 worktree 移除外,授权与停止线都不覆盖删除、覆盖、sudo、系统级或驱动安装、历史改写、强制 Git、push、丢弃未合并工作或解决合并冲突——遇到这些就停止并报告,不自行猜过去。
  • 剩下的就是目标。没有目标——问用户要;绝不从仓库里替用户猜一个。

循环之前

一次 grep 读出 .envSTAR_LANGINVOLVE 和三个模型键(§7.6、§7.7、§10.8),把每个选中的档位条目拆成模型与可选的 @<深度>;这个 harness 能指定受托者用哪个模型时,再用会话上下文里那条溯源提示给出的解析命令读一次本会话正在跑的模型(§8),下面每次启动才能判断本档是否要求不同模型或宿主支持的逐次深度。把目标变成运行可查验的判据——某个叶子要到达的 exec_status、必须存在的文件、报告里要出现的指标——开跑第一行写明判据、停止线(或说明没设)和解析出的档位。变不成判据的目标,先问清,不追。

循环

  1. 磁盘上的启动标记还在、退出文件却没出现时——本次调用留下的,或上一次留下的——先按下文「等待已启动的命令」续上等待;状态检查等退出事件之后再跑。否则经宿主原生 skill 机制运行 star-flow-status,取它唯一的下一步动作。不推进目标判据的动作不取:一行说明后留给以后的调用,改取推进判据的那个下一步。
  2. 启动被点名的那个:
    • 属于 agent 可拾起的八个——走宿主原生机制,完全按 §10.2–10.6 执行,并把本次运行解析出的档位以 involve= token 附在调用上:每次被启动的运行各自解析档位(§7.7),不带 token 的启动会退回 .env,而不是本次运行的默认值。点名的是 star-plan-executorstar-code-reviewer 且档位为 low 时,再附上 auto=unattended;该 token 传递本次调用的授权,skill 在解析目标或描述前先剥掉它。
    • 属于只许显式调用的七个——授权覆盖它:先声明一行(匹配上了什么、取了哪个目标),然后派一个子代理完整读取该 skill 在本宿主的 SKILL.md 并遵循它,本次运行解析出的档位以同一个 involve= token 写进委派简报;结束时在决策记录里留一行,匹配到什么 → 跑了什么 → 写了哪些文件(§10.5)。
    • 一条备好的红线命令——按上面的停止线规则:按下文「等待已启动的命令」启动并记录——脱离会话、写下标记、等在退出事件上而不是靠刷新——然后按交接说明收取用于查验判据的输出,再回到第 1 步;这样等不了的命令——本地没有可守的进程、也没有它的进展会触碰的文件——就结束本次运行,报告写明在等什么。守卫因审查缺失或过期而拒掉的启动不算失败的动作:对那个 run 跑 star-code-reviewer 就是下一步,启动排在它之后。被用户的停止线拦下时——打印命令、停下、报告。
    • 这个 harness 能指定受托者使用的模型或思考深度时,上面每一次启动都应用该宿主为所属档位提供的控制(§10.8),具体机制由该 harness 自己的包装文件说明;决策记录随每次启动写明档位、实际模型,以及请求的深度或 default
  3. 目标没定死就列出候选并等待(§5.2、§10.3),一次启动一个工作单元(§10.4)。low 时,auto=unattended 覆盖的动作按上面的守卫条件直接执行,不发问。其余必问确认点照问照等(§7.2);没有人能回答时——headless 运行——就停在那里报告,不替用户假设。被启动的运行递回来的问题同样分流:该档位不问的裁量题取推荐项并记录;歧义或超出无人值守授权的动作交给用户。
  4. 每次运行结束后,接着跑它点名的下一步——在本授权之下,那七个与那八个同样照跑(§10.6)——没有点名时,重跑 star-flow-status
  5. 失败过的动作不在同一目标上重试。失败换来的转交——一次修订、一次修复——本身就是下一步动作,跑一次;它也失败时,停下报告。

Read the full file on GitHub · 57 lines

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. 4d ago Changed · +1 lines 0194b24357c1
  2. 11d ago First seen · 56 lines · 0 tokens per session scan A 91787b33b10d

Subscribe to this mod's changes

star-auto.zh-CN is a command published in the GitHub repository wanghao9610/STAR (52 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,017 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.