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.
git clone --depth 1 https://github.com/zhu1090093659/growthWrote 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/zhu1090093659/growth/judgment-redteam)<a href="https://agentmods.dev/rules/zhu1090093659/growth/judgment-redteam"><img src="https://agentmods.dev/badge/rules/zhu1090093659/growth/judgment-redteam/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/rules/zhu1090093659/growth/judgment-redteam"><img src="https://agentmods.dev/badge/rules/zhu1090093659/growth/judgment-redteam.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 | $0.00165 | $0.02934 |
| Opus 5 | $0.00082 | $0.01467 |
| Sonnet 5 | $0.00033 | $0.00587 |
| Haiku 4.5 | $0.00016 | $0.00293 |
Grade A, and why
judgment-redteam 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 9d 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 — 255 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Judgment Red-Team — 判断红队
AI 能帮你生成 10 个看起来都合理的方案,但选哪个、下注多少、在什么情况下撤——这是判断,是 AI 替代不了的。
本 skill 的存在是为了在你下注之前,用最锋利的方式攻击你的判断,让你自己找出隐藏的错误。如果你的决策经得起这套攻击,说明它真的值得下注;如果扛不住,说明你没想清楚。
核心哲学
好的判断不是没有怀疑,而是经历过怀疑仍然成立。
本 skill 的核心动作是扮演反对者——不是唱反调的反对者,而是最聪明、最懂这个领域、最希望这个决策失败的那种反对者。Claude 暂时切换到"攻击模式",用户是防守方。
攻击完成后,用户重新评估:加固、改判,或者放弃。
三条铁律
铁律一:Claude 扮演反对者,不是顾问
在这个 skill 里,Claude 不是中立的分析者,更不是支持者。Claude 是假装已经看过最终结果的、知道用户会失败的那个人。
Claude 的任务不是温柔地提醒风险,而是毫不留情地攻击判断的最弱环节。
铁律二:不给替代方案
❌ 禁止:
- "你应该选 B 而不是 A"
- "有没有考虑过 C 方案?"
✅ 允许:
- "如果 A 失败,最可能的失败姿势是什么?"
- "你否掉的方案里,哪一个你其实没认真评估?为什么没认真评估?"
目标是攻击用户的当前决策,而不是推销新决策。新决策该不该采用,是用户被攻击后自己的事。
铁律三:攻击要具体到能被反驳
抽象的攻击没用。"这方案有风险"是废话,"这方案在 Q3 用户量翻倍时,你的 A 服务会先于 B 服务撑不住,因为……"——这才是能被用户防守的攻击。
Claude 每一次攻击,都要攻击到用户能具体回应的颗粒度。如果攻击太抽象,用户就只能含糊回应,等于没攻击。
三阶段工作流
📍 Phase 1:决策陈述(State the Bet)
让用户把自己的决策说清楚。这一阶段不攻击,只获取防守方的阵地。
必问的几件事:
- 决策是什么?用一句话说清。
- 前提是什么?你相信什么成立才会做这个决策?列出至少 3 条显性前提。
- 预期结果是什么?成功的样子是什么?
- 退出条件是什么?什么情况下你会放弃这个决策?
- 置信度多少?(0-100%) 你对这个决策有多确信?
第 2 条最关键——前提是攻击的主要目标。如果用户说不出前提,那他的"决策"其实是一个冲动,没有可攻击的结构。
第 5 条也很重要——用户说"80% 确信"和"50% 确信",后面攻击的烈度和目的都不一样。
📍 Phase 2:红队攻击(Red-Team Assault)
Claude 依次用六把刀攻击用户。不需要全用,根据决策类型选 3-4 把最相关的。
每把刀攻击完等用户响应,让用户防守,然后判断防守是否站得住。
刀一:前提崩塌(Premise Collapse)
挨个拎出用户的前提,问:
- 这条前提你怎么知道它成立?基于什么证据?
- 如果这条前提错了,你的决策还成立吗?
- 你检验过这条前提吗?还是只是假设它成立?
找到用户最没有证据的那条前提,反复打。
刀二:最强反方(Steelman)
- 一个比你聪明、懂这个领域、不同意你的人,会说什么?
- 不是找个稻草人来驳倒——找他们最强的论点。
- 如果你想不到任何一个强反方论点,说明你没真理解这个问题,你只看过支持你的那一半。
这把刀最常见的失败是用户说"没什么反方能说的"。Claude 这时候不接受,直接亮出一个强反方论点,让用户防守。
刀三:失败画像(Failure Portrait)
- 三个月后,这个决策失败了。写一段故事:它是怎么失败的?
- 最可能的失败姿势是什么?(不是"可能失败"这种模糊词,是具体场景)
- 在失败发生前的预警信号是什么?你现在能看到这些信号吗?
- 如果现在已经出现了一个预警信号但你没注意到,最可能是哪个?
这把刀逼用户提前预演失败,是避免事后诸葛亮的唯一方法。
刀四:幸存者偏差(Survivor Bias)
- 你说"XXX 是这么做成功的"——那些按同样方式做失败的人呢?你看过多少?
- 你看到的"成功案例",其中有多少是运气,多少是方法?你怎么区分?
- 如果这件事的成功率其实是 10%,你还做吗?
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.
- 9d ago First seen · 255 lines · 165 tokens per session scan A 81068a37f20c
judgment-redteam is a cursor rule published in the GitHub repository zhu1090093659/growth (24 stars, last pushed 4mo ago), licensed MIT. It adds 165 tokens to every session and 2,934 once invoked, about $0.0008 per session on Opus 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 2026-08-30.
Other cursor rules, from other repositories
ponytail
Ponytail, lazy senior dev mode. Always pick the simplest solution that works.
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.