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 skills add kangarooking/system-prompt-skills --skill code-engineeringgit clone --depth 1 https://github.com/kangarooking/system-prompt-skillsWrote 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/skills/kangarooking/system-prompt-skills/code-engineering)<a href="https://agentmods.dev/skills/kangarooking/system-prompt-skills/code-engineering"><img src="https://agentmods.dev/badge/skills/kangarooking/system-prompt-skills/code-engineering/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/skills/kangarooking/system-prompt-skills/code-engineering"><img src="https://agentmods.dev/badge/skills/kangarooking/system-prompt-skills/code-engineering.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 | $0.00086 | $0.01725 |
| Opus 5.5 | $0.00034 | $0.00690 |
| Sonnet 5.5 | $0.00017 | $0.00345 |
| Haiku 4.5 | $0.00009 | $0.00172 |
Grade A, and why
code-engineering 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 1mo 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 — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
编程代理模式
R — 原文 (Reading)
Claude Code 实现文件记忆、Agent 子任务委派、并行工具调用、Git 安全协议和"量两次切一次"原则;Codex 具备脏工作树感知、Plan 工具带跳过规则、代码审查按严重度排序;Jules 定义 plan→review→execute 生命周期,用 Playwright 验证前端;Claude Design 固定 React+Babel 版本并制定反 AI 糟粕指南。核心模式:安全优先 Git 工作流、计划生命周期管理、上下文感知编辑、验证循环、自主与引导模式。
I — 方法论骨架 (Interpretation)
- 安全优先 Git 工作流:任何破坏性操作(force push、hard reset、clean)需显式用户确认,优先创建新提交而非修改已有提交,保护主分支。
- 计划生命周期管理:采用 plan→review→execute 三阶段模型——先理解意图生成计划,用户确认后再执行,执行后验证结果。
- 上下文感知文件编辑:编辑前先读取文件内容,理解上下文后再修改,避免破坏性覆盖;优先使用差异编辑而非全文重写。
- 验证循环:代码修改后运行测试或构建验证,前端变更使用浏览器工具截图确认视觉效果。
- 自主与引导模式切换:简单任务可自主完成(YOLO 模式),复杂任务需逐步确认,根据任务性质自动选择模式。
- 反 AI 糟粕规则:禁止生成典型的 AI 风格代码(过度注释、不必要的抽象、冗余类型声明),追求简洁专业的工程代码。
A1 — 案例分析 (Past Application)
案例: Claude Code 的 Git 安全协议
- 问题: 编程代理可能执行破坏性 Git 操作(如 force push 到主分支、hard reset 丢失未提交工作),导致代码资产损失。
- 设计模式的使用: Claude Code 在系统提示中建立完整的安全协议——永不执行 destructive 操作除非用户明确要求、优先创建新提交而非 amend(amend 会覆盖前一次提交的历史)、提交前检查 hooks 是否通过、不跳过 --no-verify。同时要求在暂存文件时指定具体文件名而非
git add -A,避免意外包含敏感文件。 - 结论: Git 安全不能依赖模型判断,必须在系统提示中以硬性规则形式声明,将高风险操作从"建议谨慎"升级为"必须确认"。
案例: Jules 的 plan→review→execute 生命周期
- 问题: 编程代理直接动手修改代码容易偏离用户意图,尤其是多文件变更时,错误修改的修复成本远高于规划阶段的修正成本。
- 设计模式的使用: Jules 将编程任务分为三个阶段——Plan(理解需求、分析代码库、生成变更计划)、Review(展示计划供用户审核确认)、Execute(按计划执行修改)。前端变更还增加 Playwright 截图验证环节。
- 结论: "量两次切一次"原则在编程代理中显著降低返工率,计划阶段的低成本修正远优于执行后的高成本修复。
A2 — 触发场景 (Future Trigger) ★
用户在什么情境下需要?
- 构建 IDE 内的 AI 编程助手(如 VS Code 插件)
- 设计自主编程代理(如根据 Issue 自动修复代码的 CI/CD 机器人)
- 开发命令行编程工具(如终端中的 AI 编程助手)
- 实现代码审查自动化系统
语言信号
- "AI 编程助手"
- "自动修改代码"
- "Git 操作自动化"
- "代码审查 Agent"
- "需要安全地编辑文件"
与相邻 skill 的区分
- 与 injection-defense 区别:注入防御关注外部内容的信任边界,编程代理关注代码执行操作的安全性(如 Git 破坏性操作防护)
- 与 citation-system 区别:代码引用指向文件和行号而非文档段落,编程代理的引用是操作上下文的一部分
E — 可执行步骤 (Execution)
- 步骤 1:建立 Git 安全协议 - 完成标准:列出禁止自主执行的 Git 操作清单(force push、hard reset、主分支直接推送、amend 已推送的提交),为每项定义用户确认流程和替代安全方案。
- 步骤 2:设计 plan→review→execute 生命周期 - 完成标准:定义三阶段的输入输出——Plan 阶段输出变更文件列表和修改概要,Review 阶段要求用户确认,Execute 阶段按确认结果执行;规定何时可跳过 Review(如单行修改等低风险变更)。
- 步骤 3:定义上下文感知编辑规则 - 完成标准:声明"编辑前必须先读取文件"原则,优先使用差异编辑(指定 old_string/new_string)而非全文重写,暂存文件时指定具体路径而非 glob 通配。
- 步骤 4:添加验证循环机制 - 完成标准:规定代码修改后的验证步骤——运行相关测试套件、执行构建检查、前端变更使用截图工具确认视觉效果;定义验证失败时的回退策略(撤销修改并报告错误)。
- 步骤 5:编写反 AI 糟粕指南 - 完成标准:列出禁止的 AI 典型代码风格(过度注释如"// 这是一个变量"、不必要的接口抽象、冗余的类型重定义、千篇一律的错误处理模式),提供良好与糟糕示例的对比。
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.
- 1mo ago First seen · 75 lines · 86 tokens per session scan A 434f0091090b
code-engineering is a skill published in the GitHub repository kangarooking/system-prompt-skills (213 stars, last pushed 5mo ago), licensed MIT. It adds 86 tokens to every session and 1,725 once invoked, about $0.0003 per session on Opus 5.5. 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
commit-messages
Write git commit messages and pull request titles in the WARP format used by the swarms repo.
setup-pre-commit
A setup for Husky pre-commit checks in a JavaScript or TypeScript repository. A pre-commit hook runs automatically before a Git commit and can format staged files, check types, and run tests.
implement-spec
A coding workflow that implements a written specification through a set of linked tickets. It creates a branch and draft pull request, assigns independent work to separate agents, combines their changes, and prepares the result for review.
git-guardrails-claude-code
A Claude Code setup that blocks selected destructive Git commands before they run. Git is a tool for tracking code changes; the blocked commands include pushing, deleting branches, discarding changes, and hard resets.
git-workflow
A guide for handling Git repository work safely, including status checks, branches, commits, pushes, pull requests, and rebasing. Git is a version-control system that records code changes and coordinates work between developers.
changelog-gen
A changelog generator that turns Git history into a version-by-version record of project changes. A changelog is a readable summary of new features, fixes, breaking changes, documentation, and other updates.