Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计通过

develop-typescriptdevelop TypeScript 搜索

Agent Skill

develop-typescript 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。

总安装

576

周安装

24

GitHub Stars

5

下载量

192
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:develop-typescript(develop TypeScript 搜索)
来源仓库:https://github.com/yigitkonur/skills-by-yigitkonur
仓库路径:skills/develop-typescript
安装命令:
npx skills add https://github.com/yigitkonur/skills-by-yigitkonur --skill develop-typescript
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/yigitkonur/skills-by-yigitkonur --skill develop-typescript

简介

develop-typescript 编写严格类型安全的 TypeScript 代码,遵循 idiomatic 模式。

  • 涵盖 tsconfig 配置、linting 规则迁移与类型错误诊断修复。
  • 不处理 React/Vue 组件等框架特定逻辑,聚焦语言层面质量保障。
  • 适用于新项目初始化、旧 JS 迁移或类型系统审计场景。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Develop TypeScript

Write, review, or refactor framework-agnostic TypeScript with strict type safety, correct tsconfig, and idiomatic patterns.

Trigger boundary

Use this skill when:

  • writing new TypeScript code that must be strictly typed
  • reviewing existing TypeScript for type safety, anti-patterns, or missing narrowing
  • configuring or auditing tsconfig.json and linting rules
  • migrating JavaScript to TypeScript (or upgrading TS versions)
  • diagnosing and resolving type errors, inference failures, or performance issues

Do not use this skill when:

  • the task is primarily React/Vue/Angular component work (use framework-specific skills)
  • the task is exclusively about Node.js runtime APIs (use a Node.js skill)
  • the task is about build tooling only (use tooling-specific skills)

Repositories that happen to contain .tsx files are still in scope when the requested work is about TypeScript itself: tsconfig, safer typing, exported contracts, result/error patterns, or strictness audits.

Definitions

  • Load — read the file into working context so you can reference its content
  • Checked as — a cast following a runtime guard that proves the type (safe)
  • Unchecked as — a cast with no runtime proof; hides bugs (almost always wrong)
  • Block — in review mode, flag the issue and stop approving; do not auto-fix

Mode detection

Before starting, determine your mode:

SignalModeBehavior
"Write", "implement", "create", "add", "refactor", "tighten typing", "improve typing", "make TS stricter", "harden types"AuthoringWrite code, apply fixes, output files
"Review", "audit", "check", "look at"ReviewReport findings, never auto-fix, block on P0
AmbiguousAskClarify with the user before proceeding

Required workflow

Step 1 — Classify the task

Identify the primary category and optionally one adjacent category:

CategoryReference to load
Type system (generics, narrowing, inference)type-system.md
Patterns (result types, branded types, builders)patterns.md
Anti-patterns (any, ts-ignore, unsafe casts)anti-patterns.md
Error handling (Result, retry, aggregation)error-handling.md
Strict config (tsconfig, strict flags)strict-config.md
Tooling (ESLint, Biome, Prettier, CI)tooling.md
Migration (JS-to-TS, version upgrades)migration.md
Performance (compilation speed, traces)performance.md
Testing (type tests, expect-type, tsd)testing.md
Modern features (decorators, using, satisfies)modern-features.md
Steering note: Most tasks span two categories. Load the primary reference plus one adjacent. If uncertain which category fits, scan the table above for keywords that match the user request.

Step 2 — Load references

Load the reference file(s) identified in Step 1. Read the full file — do not skim.

If the task involves existing code, also load:

  • The project's tsconfig.json (compare against strict-config.md baseline)
  • The project's ESLint/Biome config (check for rule conflicts)
Steering note: When reviewing tsconfig.json, only flag moduleResolution: "node" as legacy. "node16", "nodenext", and "bundler" are all modern and correct for their contexts. Choose the baseline that matches the repo's runtime first; strictness flags are orthogonal to module mode. Do not "upgrade" node16/nodenext projects to bundler unless the user explicitly wants that runtime change. If the repo has multiple tsconfig files, start with the config that governs the files you are touching, then follow its extends chain to shared base configs. If there is no ESLint/Biome config, record that absence and skip Step 5 conflict handling rather than inventing lint policy.

Step 3 — Execute the task

Apply the patterns and rules from the loaded references.

In authoring mode:

  • Write code that follows the patterns in the reference files
  • Use satisfies for config objects, explicit return types for public functions
  • Never use bare any, @ts-ignore, or unchecked as
  • Prefer unknown + narrowing over any for dynamic data

In review mode:

  • Scan for every item in the anti-patterns.md review-mode checklist (Section: Review-Mode Scanning Checklist)
  • For each finding, classify as P0 (blocks approval) or P1 (should fix) or P2 (nit)
  • Block on: bare any in new code, @ts-ignore (should be @ts-expect-error), unchecked as, missing error narrowing
  • Flag but don't block on: style preferences, minor naming issues
Steering note: Distinguish lint warnings from type errors. Lint warnings are project-specific (respect the project's config). Type errors from tsc are universal and always blocking.

Step 4 — Audit and verify

Run these checks:

# Find unannotated exported functions (authoring mode)
grep -rnE --include="*.ts" --include="*.tsx" "^export (default )?function" . | grep -v node_modules | grep -v ": " | head -20

# Find bare 'any' usage
grep -rnE --include="*.ts" --include="*.tsx" ": any\\b" . | grep -v node_modules | head -20

# Find @ts-ignore (should be @ts-expect-error)
grep -rn --include="*.ts" --include="*.tsx" "@ts-ignore" . | grep -v node_modules

# Find unchecked type assertions
grep -rn --include="*.ts" --include="*.tsx" " as [A-Z]" . | grep -v node_modules | grep -v "// checked" | head -20

# Search for satisfies opportunities (config objects typed with annotation)
grep -rnE --include="*.ts" --include="*.tsx" "^const .* : Record<" . | grep -v node_modules | head -10
grep -rnE --include="*.ts" --include="*.tsx" "^const .* : {" . | grep -v node_modules | head -10
Steering note: The satisfies audit catches config objects that use type annotations where satisfies would preserve literal types. This is a common missed opportunity. Run these audits from the repository root, or replace . with the target package path in a monorepo. If the repo contains TSX and jsx is react-jsx, verify the React runtime/types are actually installed before treating the typecheck result as meaningful.

Step 5 — Handle conflicts

If the project's lint config contradicts skill recommendations:

  1. For new code you're writing: follow the skill's stricter rules
  2. For existing code under review: respect the project's config, flag conflicts as informational
  3. Document the conflict: "Project disables X — consider re-enabling for new code"
Steering note: Do not unilaterally re-enable lint rules. The project may have valid reasons for disabling them (e.g., gradual migration from JavaScript, third-party type issues).

Step 6 — Deliver

In authoring mode: Output the complete, compilable code. Ensure every exported function has an explicit return type annotation. For exported TSX components, follow the repo's existing convention; if none exists, prefer ReactElement via import type {ReactElement} from 'react' over relying on the global JSX.Element namespace.

In review mode: Output findings as a structured list with severity levels. Include specific line references and fix suggestions for each P0 and P1 finding.

Steering note: Always produce a deliverable — code or findings list. Never end with only commentary or advice. The user expects actionable output.

Common mistakes to avoid

MistakeWhy it's wrongWhat to do instead
Flagging node16 moduleResolution as legacyOnly bare "node" is legacyCheck strict-config.md flag table
Using any for "I'll fix it later"any disables all checking downstreamUse unknown and narrow
Ignoring cross-realm instanceof issuesObjects from iframes/workers fail instanceofUse brand checks or Symbol.hasInstance
Auto-fixing code in review modeReview should report, not modifyBlock and describe; let the author fix
Skipping the deliverable stepUser gets advice but no outputAlways output code or structured findings
Treating @ts-ignore and @ts-expect-error as equivalent@ts-ignore suppresses silently foreverAlways prefer @ts-expect-error — it alerts when the error is fixed

Reference routing

FileLoad when
type-system.mdWorking with generics, conditional types, type guards, narrowing, variance
patterns.mdImplementing Result types, branded types, builders, middleware, retry logic
anti-patterns.mdReviewing code for unsafe patterns or scanning for anti-pattern violations
error-handling.mdImplementing error handling, retry logic, error aggregation, or Result types
strict-config.mdConfiguring or auditing tsconfig.json and strict mode flags
tooling.mdSetting up or choosing between ESLint, Biome, Prettier, CI pipelines
migration.mdMigrating JS to TS, upgrading TS versions, handling circular dependencies
performance.mdDiagnosing slow compilation, reading traces, optimizing type-check speed
testing.mdWriting type-level tests, choosing test strategies, testing generics
modern-features.mdUsing decorators, using/Symbol.dispose, satisfies, inferred predicates

Guardrails

  • Never introduce any — use unknown and narrow
  • Never use @ts-ignore — use @ts-expect-error with explanation
  • Never use unchecked as — always guard first, then cast
  • Always add explicit return types to exported functions
  • Always verify tsconfig against strict-config.md baseline before suggesting config changes
  • In review mode: report findings, never silently modify code

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

33.95%
按下载量换算65

Claude

31.87%
按下载量换算61

Cursor

16.32%
按下载量换算31

Gemini CLI

9.13%
按下载量换算18

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills