Token导航 LogoToken导航

Agent Skill · 研究检索

tt-polish(tt 抛光)

tt-polish 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

tt-polish 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,暂无第三方扫描结果,按安装量在研究检索类的 73,675 个 Skill 里排第 65,745。

总安装 275周安装 11GitHub Stars 公开资料未说明下载 89核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:tt-polish(tt 抛光)
来源仓库:https://github.com/biasia/tt-impeccable
仓库路径:skills/tt-polish
安装命令:
npx skills add https://github.com/biasia/tt-impeccable --skill tt-polish
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/biasia/tt-impeccable --skill tt-polish

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

tt-polish 用于查找、检索和筛选相关信息。适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词或任务场景快速定位候选结果。通过 npx skills add 命令从指定仓库安装并使用。安装前建议确认权限范围和维护状态,注意是否触发联网或文件读写操作。

tt-polish 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/tt-polish

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/biasia/tt-impeccable --skill tt-polish;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

tt-polish 怎么安装?

在终端执行 npx skills add https://github.com/biasia/tt-impeccable --skill tt-polish。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

tt-polish 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

tt-polish 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

tt-polish 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

tt-polish 用的人多吗?

公开数据是累计安装 275 次,近一周安装 11 次,下载 89 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

tt-polish 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 要求有安全扫描结论才肯安装的团队。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
tt-polish

tt 抛光 · 本页

275—只读未明确未收录

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Polish is a two-phase pass, not a builder. It assumes TUX token compliance is already mostly in place — if it isn't, run /tt-impeccable extract first.

Phase 1 — Visual Elevation: adds warmth, depth, and richness that survive token compliance but are not prescribed by the core spec. Phase 2 — Compliance Sweep: catches token drift, missing states, anti-patterns, and copy issues.


Context Check

  1. Read .tt-impeccable.md in project root if it exists — fidelity tier and design context are there.
  2. If absent, infer tier from codebase: mobile-first 390px = Strict TUX-native; dark #000000 desktop = TT Style.
  3. Only ask if the tier is genuinely ambiguous.

Note the fidelity tier: Phase 1 rules apply primarily to Strict TUX-native light-theme surfaces. Phase 2 applies to both tiers.


Phase 1: Visual Elevation

Load reference/polish-rules.md and apply all 13 rules systematically.

Rule summary (refer to the rules file for full CSS and constraints):

#If… (condition)→ Then (action)
1Page has a prominent heading (font-size > 17px)Left-align it, add a subtitle line below, place a radial-gradient glow at top — height = heading's top offset + 20px so gradient is fully transparent by the time the heading begins; color from Ambient Glow presets (rules file); skip on utility/settings screens
2Largest font-size on page < 20pxSet one heading to 24–32px (scale to content density); compress smallest body text to 12px
3Page horizontal margin < 20px, or element gaps feel tightSet page padding to 20px; inter-group vertical gaps to 24–32px
4Card or icon list cell presentReplace every icon with a contextual emoji; set background to ~5% tint of dominant hue using oklch() (keeps color vivid)
5Card or cell border-radius < 12pxUpgrade: L1 nested → 14px, L2 overlay/sheet → 26px
62–5 emoji list cells presentShorten label text, foreground the emoji, reflow as a 2-column card grid
7Same-type cards in a gridEnforce a consistent internal layout system across all cards: same emoji size, same title position, same description line count, same padding — no card should look structurally different from its siblings
8Divider lines presentReplace with fading-edge gradient; attach to scroll wrapper (not chrome elements)
9≥3 distinct non-neutral colors on screenSwitch CTA button to black (#000000) with white label
10Icon sizes other than 24px or 16px presentSnap large icons → 24px, small icons → 16px; enforce one size per component
11Solid non-neutral fill (red, blue, etc.) > 30px wide/tall with icon insideRemove the fill container; promote the associated title to 24px/700 dominant typography
12Entrance animations absentAdd: cards fadeUp staggered 40ms, sheets fadeScale 220ms, chips fadeIn staggered 30ms; disable translate/scale under prefers-reduced-motion
13Group/section label is visually stronger than item titles inside itDemote label to 13px/600/text-2; item titles stay at 15–17px/700/text-1-display — the label is wayfinding, not a headline

Phase 1 rules apply to Strict TUX-native light-theme surfaces. For TT Style surfaces, Rules 2, 3, 5, 10, 11 still apply; Rules 1, 7, 12 are adapted (see rules file for TT Style notes).


Phase 2: Compliance Sweep

2a — Token Drift Scan

Read source files and search for:

  • border:, border-top:, border-bottom:, border-left:, border-right: — for each hit: (a) could a background color change replace it? (b) Is it compensating for wrong surface contrast? Flag P1/P2 as appropriate.
  • :hover rules containing border-color or adding a border — convert to background/opacity changes (P2)
  • backdrop-filter — confirm glassmorphism is explicitly confirmed in context; if not, remove (P1)
  • letter-spacing non-zero on text below 72px — remove (P2)
  • rgba(254,44,85,*) at any alpha on non-CTA/non-interactive elements — BAN 7 violation (P1)
  • Accent color (brand red, teal) on elements without interactive semantics (no onClick/href/role) — replace with neutral (P1)
  • Hard-coded hex or rgba values not going through var(--tux-*) — flag all (P2)
  • Spacing values not in the TUX 4px grid (4/8/12/16/20/24/32) — flag off-grid values (P2)
  • border-radius values not in the TUX tables — flag (P2)
  • box-shadow values that don't match the 6 TUX levels — flag (P2)
  • Font other than TikTok Sans / system-ui — flag (P1)
  • overflow: hidden on text containers with line-height below 1.0em — add descender padding fix (P2)

2b — Interaction States

Confirm every interactive element has all applicable states:

  • Default — resting appearance
  • Active / pressed — opacity drop or background shift on touch
  • Focus-visible — keyboard/accessibility ring; never outline: none without replacement
  • Disabled — visually de-emphasized; pointer-events: none
  • Loading — if async: spinner, skeleton, or progress; element must not look tappable while loading
  • Error — if element can fail: validation state visible
  • Success — if action has meaningful completion: confirmation or state change

Touch target minimum: 44×44px on every tappable element.

2c — Entrance Animations (from Phase 1 Rule 12)

  • All elements have entrance animations — no element snaps into existence at full opacity
  • Cards/items use staggered fadeUp; floats use fadeScale; chips use staggered fadeIn
  • prefers-reduced-motion: translate/scale animations disabled; opacity-only or none

Phase 1 Checklist (before presenting)

Gradients & Glow

  • Page has heading > 17px → gradient applied; fades fully transparent before 220px
  • Maximum one glow source per page

Typography

  • Whole-page: largest text ≥ 24px, smallest ≤ 12px
  • Within each module: at least one 700-weight dominant and one 400-weight/12px recessive element
  • Adjacent tiers differ by ≥ 2 steps on the TUX scale

Spacing

  • Page horizontal padding is 20px
  • Inter-group vertical gaps are 24–32px
  • Card internal padding is 20–24px

Cards & Emoji

  • All SVG/system icons in cards and list cells replaced with contextual emoji
  • Card/cell backgrounds use oklch() tint at ~5% — not flat rgba()
  • Card label/heading color carries the card's hue
  • 2–5 emoji list cells reflowed as 2-column card grid
  • White-on-white: card gets tinted background

Fills & Color Restraint

  • No solid non-neutral fill > 30px containing an icon — removed and replaced with dominant title
  • CTA: ≥3 distinct non-neutral colors on screen → CTA is black

Surface & Structure

  • Radius: L1 nested ≥ 14px; L2 overlay = 26px
  • Same-type card grid: all cards share identical internal structure (emoji size, title position, description lines, padding)

Fading Edges

  • Dividers replaced with fading-edge gradient attached to scroll wrapper
  • Fade color matches scroll surface background
  • Fade height ≥ 32px

Icons

  • Icon sizes snapped to 24px (large) or 16px (small); one size per component

Typography Hierarchy

  • Group/section labels are ≤13px/600/text-2 — weaker than item titles inside them

Animations

  • Cards: fadeUp staggered 40ms per item
  • Sheets/modals: fadeScale 220ms
  • Chips: fadeIn staggered 30ms
  • prefers-reduced-motion: translate/scale disabled

Final Verification

  1. Use it yourself — interact with every state and flow
  2. Run /tt-audit — confirm health score improved
  3. Re-run /tt-critique — verify anti-pattern verdict is clean

Polish is complete when the interface is indistinguishable from a screen inside the TikTok app — not because it has the right tokens, but because it has the right *feel*.

平台分布

Codex

33.75%
按下载量换算30

Claude

29.36%
按下载量换算26

Cursor

18.6%
按下载量换算17

Gemini CLI

9.06%
按下载量换算8

安全审计

暂无安全审计结果可展示。

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills