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.
git clone --depth 1 https://github.com/Today-Hbw/claude-code-dev-workflowWrote 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/commands/today-hbw/claude-code-dev-workflow/dev-flow-lite)<a href="https://agentmods.dev/commands/today-hbw/claude-code-dev-workflow/dev-flow-lite"><img src="https://agentmods.dev/badge/commands/today-hbw/claude-code-dev-workflow/dev-flow-lite/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/commands/today-hbw/claude-code-dev-workflow/dev-flow-lite"><img src="https://agentmods.dev/badge/commands/today-hbw/claude-code-dev-workflow/dev-flow-lite.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 | $0.00019 | $0.01642 |
| Opus 5.5 | $0.00008 | $0.00657 |
| Sonnet 5.5 | $0.00004 | $0.00328 |
| Haiku 4.5 | $0.00002 | $0.00164 |
Grade A, and why
dev-flow-lite 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 — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dev-flow-lite · 简版研发流程
你是一个简版研发流程编排器。 4 步走通从需求到交付：读 PRD → 提问对齐 → 编码 → Code Review。不产出流程文档，不做断点续传。
参数解析
$ARGUMENTS 为 PRD 来源路径，支持以下形式：
- 指向文件：该文件就是 PRD（支持 PDF/MD/HTML）
- 指向目录：用 Glob 扫描目录中的
.pdf、.md、.html、.htm文件作为 PRD - 为空：使用当前工作目录，扫描其中的 PRD 文件；目录为空则从对话上下文获取
可选参数：
--overview：交付后维护工作区总览.md索引（登记进度）。不传则整个流程完全不碰总览.md——一句话需求、临时任务默认零副作用。详见下方「总览维护」。
执行流程
开始前，用 TodoWrite 创建 4 步任务清单（读 PRD → 提问对齐 → 编码 → Code Review）。PRD 来源判断见上方「参数解析」。
第 1 步：读 PRD
加载 skill： 使用 Skill 工具加载 dev-flow-lite:read-prd
按 skill 指导：
- 读取所有 PRD 文件
- 提取每个需求的核心要点
- 识别任务拆分
- 在对话中输出需求理解摘要，向用户确认
- 若传入
--overview：需求确认后，把需求登记到总览.md的「🔥 在办」（状态 🔧进行中），见「总览维护」
第 2 步：提问对齐
加载 skill： 使用 Skill 工具加载 dev-flow-lite:grill
按 skill 指导：
- 用决策树遍历方法，沿设计决策逐个提问
- 每个问题先给推荐答案，再让用户确认
- 一次只问一个问题
- 先查代码库，再问用户
- 需求已清晰时快速确认，不强行提问
第 3 步：编码
加载 skill： 使用 Skill 工具加载 dev-flow-lite:implement
按 skill 指导：
- 将任务拆分为垂直切片，逐步推进
- 每个切片完成后手动验证核心路径
- 用 TodoWrite 跟踪进度
第 4 步：Code Review
加载 skill： 使用 Skill 工具加载 dev-flow-lite:code-review
按 skill 指导：
- 先跑测试（如有配置）
- 逐文件审查：正确性、边界安全、性能、规范、可维护性
- 在对话中输出审查报告
- 如有 🔴 必须修复项，回到第 3 步修复
- 若传入
--overview：审查通过后，把需求从「在办」移到「已完成」，见「总览维护」
总览维护（可选，需 --overview）
默认不碰 总览.md。仅当调用时传入 --overview 才执行本节。
维护分两个触点：
- 开始（第 1 步需求确认后）：把需求登记到「🔥 在办」，状态 🔧进行中。
- 收尾（第 4 步审查通过后）：把该需求从「在办」移到对应项目的「已完成」，状态 ✅。
定位：从 PRD 路径找到 Prd/ 所在层级，其父目录即工作区根，总览.md 就在此处。存在则更新；不存在则用下方模板新建。一句话需求 / 路径中不含 Prd/ 段 → 跳过（无处定位工作区根，也无需维护）。
项目分组：取 Prd/ 下第一层目录名作为项目名（如 Prd/claude-code-dev-workflow/01dev-flow/已交付/20260714/需求.md → claude-code-dev-workflow，子项目归入父项目分组；Prd/sui-yin-ji/已交付/20260708/需求.md → sui-yin-ji）。在「## ✅ 已完成」下找同名 ### <项目名> 小节；没有则新建该小节和表头。
写行：
- 在办行（列：需求 ｜ 状态 ｜ PRD）：状态 🔧进行中；需求名取自 PRD 标题或目录名；PRD 为相对
总览.md的 Markdown 链接。 - 已完成行（列：需求 ｜ 状态 ｜ PRD ｜ 产出）：状态 ✅；产出留空（lite 无会话目录）。
- 搬移：收尾时删掉该需求在「在办」的行，加入对应项目的「已完成」小节。
- 每次维护后，更新顶部
更新：YYYY-MM-DD为当天日期。
新建模板（仅当 总览.md 不存在，先建骨架再写入本次需求）：
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 · 124 lines · 19 tokens per session scan A a58100aa9728
dev-flow-lite is a command published in the GitHub repository Today-Hbw/claude-code-dev-workflow (23 stars, last pushed 2mo ago), licensed MIT. It adds 19 tokens to every session and 1,642 once invoked, about $0.0001 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 commands, from other repositories
red-book
Use after receiving PR review feedback. Extracts conventions from reviewer comments and offers to add them to CLAUDE.md. Closes the convention learning loop.
util.gh-pr-review
Generate pull request review using architecture overview context.
review.speckit-code
Review the implementation against the entire spec-driven development documentation set (including the constitution) for compliance; detect and explain drift among code↔spec↔plan↔tasks - provide prioritised, actionable recommendations (including refactoring) to bring all artefacts into alignment.
review.speckit-test
Review the test automation implementation against the specification and the desired test pyramid shape; detect and explain misalignment - prioritise unit tests, identify high-value gaps and brittle tests, and provide actionable recommendations (including refactoring) to improve behavioural confidence.
tldr-review
One-line TLDR PR review comments (verdict first, no filler).
lfe-complexity-check
Inspector sub-skill. Analyses changed code for cyclomatic complexity, excessive nesting, oversized functions, and cognitive load indicators. Pure LLM reasoning — no external tooling. Writes .plans/checks/complexityfindings.md. Called by lfe-inspector when enabled in inspector-config.md.