菜单
Use when implementing any feature or bugfix, auditing an implementation against its plan, or resolving a bug. This is the top-level programming skill that routes to the correct workflow. Triggers on any programming task that isn't purely reading or researching code.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/StrongAI/claude-skills-programming --skill programming命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | programming |
| description | Use when implementing any feature or bugfix, auditing an implementation against its plan, or resolving a bug. This is the top-level programming skill that routes to the correct workflow. Triggers on any programming task that isn't purely reading or researching code. |
Three paths. Identify which one applies and follow it.
A new feature, enhancement, or planned change.
/think — brainstorm the design, explore approaches, get approval/write-plan — create a detailed implementation plan from the design/dispatch-programmers — execute the plan, dispatching subagents per task
/implement — the TDD loop with audit agent per round/wrapping-up-programming — verify everything works, commitEach task dispatched in step 3 MUST load the appropriate language/framework skill:
| Language/Framework | Skill | Also load |
|---|---|---|
| Swift | /programming-swift | /swift-tdd for test-driven work |
| SwiftUI | /programming-swift-ui | /programming-swift always |
| TypeScript / React / Next.js | /programming-typescript | |
| C++ | /programming-cpp | |
| PostgreSQL | /programming-postgres | |
| Concurrent code (any language) | /programming-concurrency | language skill |
| Reverse engineering / binaries | /decompiling |
If the task spans multiple languages or frameworks, load all applicable skills. The language skill defines patterns, conventions, and anti-patterns that the subagent must follow.
Verifying that what was built matches what was planned.
/audit-code — clean agent reads the plan, reads every changed file, produces a findings reportThis runs AFTER implementation, not during. The auditor is a separate agent with no implementation context.
Something is broken and needs to be fixed.
/solving-bugs — debugger-first root cause investigation, fix the bug/wrapping-up-programming — verify the fix works, commitDo NOT use Path 1 for bugs. Bugs need a debugger, not a design phase.
| Signal | Path |
|---|---|
| "Implement", "add", "build", "create", new feature | Path 1: Implementing |
| "Audit", "verify", "check against plan", "review what was built" | Path 2: Auditing |
| "Fix", "broken", "bug", "crash", "not working", "wrong behavior" | Path 3: Bug |
If unclear, ask the user which path applies.