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 skills/yilecoding/fastapi-react-admin/ctxnpx skills add yilecoding/fastapi-react-admin --skill ctxgit clone --depth 1 https://github.com/yilecoding/fastapi-react-adminWhat 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 | $0.00067 | $0.01247 |
| Opus 5 | $0.00034 | $0.00624 |
| Sonnet 5 | $0.00013 | $0.00249 |
| Haiku 4.5 | $0.00007 | $0.00125 |
Grade A, and why
ctx 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 yesterday.
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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
往上下文文档里加一条结论
这个仓库的文档是踩坑记录，不是使用手册。判据只有一条：
违反了会坏，而且多数是静默地坏。
不满足这条的不要写进去 —— 风格偏好、能从代码直接读出来的事实、 一次性的调试过程，都不属于这里。写进去只会稀释真正重要的那些条目。
一、先定位写到哪一份
写到离代码最近的那一份，不要往根 CLAUDE.md 堆。根文件在会话开始时
全量加载，而分册是读到那个目录的文件时才惰性加载 —— 写在分册里不占别人的预算。
| 结论关于… | 写进 |
|---|---|
| 通用组件、尺寸覆盖失效、组件选型 | packages/ui/AGENTS.md |
| 查询区字段 / 运算符 / 出参映射 | packages/ui/src/components/query-bar/AGENTS.md |
| 富文本、内联图、公开子树 | packages/ui/src/components/rich-text/AGENTS.md |
| 多页签、偏好、侧边栏、外壳滚动 | packages/platform/src/shell/AGENTS.md |
| 页面模板、主从页/监控页滚动骨架 | packages/platform/src/pages/AGENTS.md |
| 设置屏骨架本身 | packages/platform/src/pages/_shared/AGENTS.md |
| 参数配置那种设置屏 | packages/platform/src/pages/config/AGENTS.md |
| 菜单管理、死链判定 | packages/platform/src/pages/menu/AGENTS.md |
| 文件上传 / 预览 / 附件 | packages/platform/src/pages/file/AGENTS.md |
| 文案、语言包、两个校验脚本 | packages/i18n/AGENTS.md |
| 后端模型 / 接口 / SQL / pytest | apps/api/AGENTS.md |
| 定时任务、Celery、调度、删除类任务 | apps/api/backend/app/task/AGENTS.md |
| Playwright E2E | apps/web/e2e/AGENTS.md |
| 跨模块、任何一层都可能违反的 | 根 CLAUDE.md 的「硬纪律」，并给个编号 |
新目录还没有分册就新建一份：<目录>/AGENTS.md 真身 +
ln -s AGENTS.md <目录>/CLAUDE.md（Claude Code 只认 CLAUDE.md，
其余 agent 工具认 AGENTS.md）。
二、按这个仓库的文体写
一条合格的结论有四件东西，缺了「症状」和「实测」的条目没人敢信：
- 症状 —— 界面上/日志里看到的是什么。写得越具体越好， 下一个人是靠症状检索到这条的，不是靠根因
- 根因 —— 为什么会这样
- 修法 —— 改哪里，以及为什么不能改在别处
- 实测证据 —— 具体数字、具体报错原文。「实测」两个字是这份文档的信用基础
排版沿用现有习惯：真的会让人损失时间的用 🔴，需要留意的用 ⚠️，
对照关系用表格，反例用 ❌ / 正例用 ✅。
反面例子（不要这样写）：
- tooltip 要正确使用 TooltipTrigger
正面例子：
🔴
TooltipTrigger render={<span className="contents" />}会把气泡甩到视口左上角。display:contents不生成布局盒，getBoundingClientRect()返回全 0， 而 Base UI 拿这个 span 当定位参照。实测（@base-ui/[email protected]）： 按钮在{x:308,y:308}，气泡落在{x:0,y:4}。 修法是把render直接指向按钮本身（TooltipTrigger render={trigger} />）， 三层嵌套TooltipTrigger → DropdownMenuTrigger → Button也成立，实测已贴合。
三、收尾必做
pnpm ctx:check
它核对「指向的东西还在不在」（死路径 / 死链接 / 死脚本 / 死 testid / 行数预算）。 它不校验文字对不对 —— 那还是要人读。
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.
- yesterday First seen · 82 lines · 67 tokens per session scan A dd0e16470b51
ctx is a skill published in the GitHub repository yilecoding/fastapi-react-admin (11 stars, last pushed 2d ago), licensed MIT. It adds 67 tokens to every session and 1,247 once invoked, about $0.0003 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 skills, from other repositories
raqam
Teaches the raqam React number-input library — headless NumberField compound components, useNumberFieldState/useNumberField hooks, Intl presets, locale plugins (fa/ar/bn/hi/th), raqam/server formatting, validation, ScrubArea, accessibility, and form integration. Use when building or debugging currency/percent/unit…
codexspec:distill
Distill reusable, cross-feature knowledge from an interaction into the project profile.
uzbek-humanizer
Use when writing, rewriting, reviewing, or localizing Uzbek Latin copy (UI strings, quiz stems, marketing, chat, bank/OTP, consent, health soft UI). Owns natural-Uzbek rewrite quality: calques, soft synonyms, Siz register, oʻ/gʻ orthography. Trigger for o'zbekcha, humanize Uzbek, matnni tabiiylashtirish…
gitee-expert
You have access to Gitee (gitee.com), China's largest GitHub-style platform. This skill teaches the effective workflows for discovery, intel, translation and webhook monitoring.
chinese-markdown-style
House style for Chinese Markdown docs, and how to enforce it with the cndocstyle package from cn-doc-style-guide.
frontend-feature
Build a new page, view, or data-driven feature in the Next.js frontend. Use when adding a route under the dashboard/marketing area, wiring UI to a backend endpoint, adding client state, or creating a localized page. Covers App Router, data fetching, Zustand stores, and i18n.