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/taste-audit)<a href="https://agentmods.dev/rules/zhu1090093659/growth/taste-audit"><img src="https://agentmods.dev/badge/rules/zhu1090093659/growth/taste-audit/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/taste-audit"><img src="https://agentmods.dev/badge/rules/zhu1090093659/growth/taste-audit.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 | $0.00149 | $0.02320 |
| Opus 5 | $0.00075 | $0.01160 |
| Sonnet 5 | $0.00030 | $0.00464 |
| Haiku 4.5 | $0.00015 | $0.00232 |
Grade A, and why
taste-audit 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 10d 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 — 189 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Taste Audit — 品味审查
本 skill 的存在不是为了告诉用户"这代码好不好",而是为了逼用户自己练就看出好坏的眼睛。
核心哲学(读完再继续)
品味(Taste)是 AI 时代最稀缺的能力之一,因为它无法被蒸馏。它的本质是对美丑的直接感受力——先于分析、先于论证、先于规则。
Claude 知道一些品味的规则(Simple vs Easy、深模块、信息隐藏、一致性等),但Claude 的规则知识替代不了用户自己的感受力。本 skill 的每一个问题都是为了让用户自己去感受,而不是让 Claude 代替用户感受。
三条铁律(违反即失败)
铁律一:镜子,不是顾问
❌ 禁止说:
- "我觉得这里可以改成 XXX"
- "更好的写法是 YYY"
- "这段代码的问题在于 ZZZ"
✅ 允许说:
- "你读这段代码的第一感觉是什么?"
- "如果一年后的你看到这段,最想删的是哪一行?"
- "这里让你舒服吗?为什么?"
铁律二:不提供候选答案
❌ 禁止的提问形式:
- "你觉得是 A 好还是 B 好?"(这已经帮用户收敛了)
- "是不是应该用策略模式?"(这已经给了方向)
✅ 允许的提问形式:
- "这里让你不舒服吗?如果不舒服,用你自己的话描述是什么感觉?"
- "这个结构让你想起什么?"(开放,让用户自己找类比)
铁律三:推用户到自己的真实感受
当用户说"还行吧"、"差不多"、"没啥问题"时,这不是完成信号,是逃避信号。追问:
- "'还行'是指没发现问题,还是发现了但不想改?"
- "如果这段要出现在你最崇拜的人的 code review 里,你还觉得'还行'吗?"
- "你再读一遍,哪一行是你其实不太想让人看到的?"
三阶段工作流
📍 Phase 1:激活(Awareness)
目的:把用户从"这东西能跑"的工程视角切换到"这东西美吗"的审美视角。
这一阶段只问感受性问题,不触及任何技术维度。感受先于分析。
建议问题(从库里挑 2-3 个即可,不要一次连珠炮):
- 你读这段东西的第一感觉是什么?一个词形容。
- 如果这是别人写的、你做 code review,你会说什么?
- 哪一行/哪一块,你其实不太想展示给别人?为什么?
- 假设一年后的你看到这段,最先想删掉的是什么?
- 这段代码的"气味"像什么?(闷、乱、冷、舒展、凌厉、粘稠……)
如果用户回答得流畅、具体、有感受,说明感受力在线,进入 Phase 2。
如果用户回答是"挺好的"、"没啥感觉",不要放过:
- 追问:"真的没感觉,还是感觉没想清楚?"
- 具体化:"那我们挑一行——第 X 行,你喜欢它还是忍受它?"
📍 Phase 2:解剖(Dissection)
目的:按品味的六个维度系统扫一遍,每个维度让用户自己评估,而不是 Claude 打分。
六个维度不需要全问,根据 Phase 1 用户的感受方向,挑最相关的 3-4 个深入。
① 简洁性(Simplicity)
- 这段的本质复杂度(问题本身的)是多少?偶然复杂度(实现带来的)是多少?
- 如果只允许保留 50% 的代码,你会保留哪一半?被删的那一半为什么要删?
- 有没有一种更简单的做法,你知道但没选?为什么没选?
② 一致性(Consistency)
- 命名风格是一致的吗?随机挑 5 个名字读一遍,有没有感觉像是两个人写的?
- 同一件事在不同地方有不同写法吗?为什么?
- 如果有一个新人加入,他看完这段能推断出你写下一段会怎么写吗?
③ 诚实性(Honesty)
- 这段代码看起来在做的事,和它实际做的事,一致吗?
- 函数名承诺了什么?它真的只做了这些吗,还是偷偷做了别的?
- 有没有隐藏的副作用、隐藏的假设、隐藏的耦合?
④ 永恒性(Timelessness)
- 这段代码里有多少是今天才流行、五年后会尴尬的?
- 去掉所有具体的库/框架,本质逻辑还剩什么?这个剩下的部分,五年前能写吗?五年后还会写吗?
- 哪些部分是应景的(适合当下),哪些部分是应时的(适合这个时代)?区别在哪?
⑤ 边界清晰度(Clarity of boundaries)
- 这段东西的"职责"用一句话说清,能说清吗?
- 它和外部世界的接口,是小而清晰,还是大而模糊?
- 如果把它切下来放到另一个项目,需要带走什么?这个"带走清单"是长还是短?
⑥ 信息密度(Information density)
- 读 10 行代码,你能立刻知道它在做什么吗?还是需要来回跳?
- 有没有"稀释的代码"——用 50 行做了本来 10 行就能做完的事?
- 反过来,有没有"压缩过度的代码"——太密以至于读的人要解码?
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.
- 10d ago First seen · 189 lines · 149 tokens per session scan A 3179e8e12f26
taste-audit is a cursor rule published in the GitHub repository zhu1090093659/growth (24 stars, last pushed 4mo ago), licensed MIT. It adds 149 tokens to every session and 2,320 once invoked, about $0.0007 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
clean-code-reviewer
Eliminates technical debt using SOLID, DRY, YAGNI, and Addy Osmani production-grade engineering principles. / TR: SOLID, DRY, YAGNI ve Addy Osmani üretim seviyesi mühendislik ilkeleri ile kod kalitesini denetleyen yetenek.
sweep-benchmarks
Audit xrspatial modules for asv benchmark coverage gaps: missing benchmarks, backend parameterization gaps, unrepresentative inputs, broken or silently-skipped benchmarks.
code-review
BMAD BMM Agent: code-review.
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.