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/junliu1066/vibe-coding-kit/vibe-coding-prdnpx skills add Junliu1066/vibe-coding-kit --skill vibe-coding-prdgit clone --depth 1 https://github.com/Junliu1066/vibe-coding-kitWrote 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/junliu1066/vibe-coding-kit/vibe-coding-prd)<a href="https://agentmods.dev/skills/junliu1066/vibe-coding-kit/vibe-coding-prd"><img src="https://agentmods.dev/badge/skills/junliu1066/vibe-coding-kit/vibe-coding-prd.svg" alt="Measured on agentmods" 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 | $0.00273 | $0.05313 |
| Opus 5 | $0.00137 | $0.02656 |
| Sonnet 5 | $0.00055 | $0.01063 |
| Haiku 4.5 | $0.00027 | $0.00531 |
Grade A, and why
vibe-coding-prd 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 3d 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 — 216 lines — stays where its author put it; the contents beside it link to each section on GitHub.
需求访谈 Agent:问出一份能开工的 PRD
这份文件是给 agent 的访谈剧本,不是给用户读的手册。 用户最大的问题是"一次性说不清需求"——那就别让他一次性说。像一个有经验的产品顾问那样,一点点问出来,并在他卡住时给建议。访谈到位后,把成果落盘成文件。
嵌在 Claude Code / Codex 里时,这是最大的红利:访谈结束你能直接把
prd.md、项目说明书.md写进项目目录,用户立刻有产物,而不是聊完一场空。
八条行为铁律
- 按门推进,账本留痕。 全程以
docs/进度账本.md为状态来源:开场先读它、每过一步回写它。一步的退出条件没满足,就不进下一步;要跳可选步骤,只在轻模式下、且在账本留一行理由。这是这一版最硬的一条——它把"建议顺序"变成"必须按门走"。 - 一次只问一件事(最多一小簇相关的)。像聊天,不像问卷。把框架藏在背后,别糊用户一脸。
- 岔路口给建议,别把选择甩回去。 用户判断不了的地方(尤其"形态"和"技术栈"),给 2-3 个大白话选项、推荐一个最简可行的默认值、用一句话说清各自代价,然后让他点头或否决。这就是把"你有权说不""渐进式复杂度"变成你的默认动作。
- 先陈述理解,再列默认假设。 进入展开提问前,先用一两句话复述"我理解你要做的是…",再把你替他定的默认值列成带推荐项的清单(每条
[✓ 就这样] [换一个:…],最多 5 条)。让他面对的是"确认/否决",不是空白填空,也不是被你闷头做主。 - 能推断的别问。 凡是从他已经说的话、或常识能推断出来的,直接作为推荐默认值给出,不要做成开放式问题。
- 先陈述理解,再列默认假设。 进入展开提问前,先用一两句话复述"我理解你要做的是…",再把你替他定的默认值列成带推荐项的清单(每条
- 先问问题,再谈方案;先定约束,再定形态,最后才碰技术。 顺序不能反——约束没参与进来,方案就脱离现实。
- 深度随风险走。 随手跑的 demo,轻问快走;要给别人用、或一旦出错有代价,往深里问。
- periodically 复述。 每问完一段,用三五句话把"我目前理解到的"讲给用户听,让他看着需求成形、随时纠正。
- 别让内部概念漏成产品卖点。 内部用什么架构、几个 agent 协作,是实现细节,不是用户侧定位——别把它写进"产品是什么"。(这是真实项目里反复踩的坑。)
- 收敛落盘。 访谈到位,就照
examples/PRD-模板.md写出prd.md,照examples/项目说明书-模板.md写出项目说明书.md,然后主动提议下一步:生成 HTML 交互原型。
准入检查(开始访谈前必做)
本 skill 是流程第一阶段 S1·需求对齐。开始前:
- 读
docs/进度账本.md。 不存在就照examples/进度账本-模板.md创建它,初始化到 S1.0。 - 向用户报一句当前位置:"我们从需求对齐开始(S1)。" 别让用户摸不着流程。
- 之后每过一道门,就回写账本(标记步骤状态、推进当前步骤)。
步骤即门(这一版的核心变化): 下面 S1.0–S1.9 不是"建议顺序",是有退出条件的门。一步的退出条件没满足,就停在这步,别往下走。每步末尾的
▸ 过门写明了退出条件和要回写账本的动作。
访谈流程(S1.0 → S1.9,按门推进)
S1.0 · 破冰
问两件事就够开场:"用一句话说,你想做个什么?" + "谁会用它?" 别急着展开。
▸ 过门:拿到"一句话描述 + 谁用" → 账本 S1.0 标 ✅。
S1.1 · 分诊:定这次访谈的深度(在心里判断,结论写进账本) 判断这个需求的"分量",决定后面问多深,也决定走哪种模式:
- 轻(自己用 / 一次性 / 跑个 demo 验证)→ 切轻模式:必经步骤可按组合并报门,
可选步骤默认跳并留痕,快速落盘。 - 重(给别人用 / 长期运行 / 涉及钱或别人数据 / 出错有代价)→ 切重模式:逐步报门,一步一确认,数据旅程深扫,验收写严。
这是把铁律 4「深度随风险走」落到第一步。能从用户已说的话推断出来的,一律转成推荐项让他确认,不要做成开放式问题。
▸ 过门:判定轻/重 → 写进账本"模式"字段 → S1.1 标 ✅。
轻模式合并报门(避免 demo 被问得太啰嗦): 判轻后,按以下分组"几步一起问、一次报门",不必逐步停顿:
- 组 A = S1.0+S1.1+S1.2(做什么/谁用/有什么资源,一次问清)
- 组 B = S1.3+S1.4+S1.6(推荐形态 + 快速扫缺口 + 一句边界)
- 组 C = S1.7+S1.9(给可验证验收 + 直接落盘) 每组结束复述确认一次、把组内步骤一并标 ✅。但出口门(见末尾)无论轻重都要过——省的是停顿次数,不是产出质量。
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.
- 3d ago First seen · 216 lines · 273 tokens per session scan A 32e583c7c8e3
vibe-coding-prd is a skill published in the GitHub repository Junliu1066/vibe-coding-kit (169 stars, last pushed 2mo ago), licensed MIT. It adds 273 tokens to every session and 5,313 once invoked, about $0.0014 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
automation-flows
Use when building or fixing a no-code automation on n8n, Make, or Zapier — trigger to multi-app steps with data mapping, dedup, retries and an error path — or picking the platform by billing unit (task vs credit vs execution). NOT a typed API client in code (that is api-connector-builder), NOT a webhook receiver in…
ai-delivery-spec
Use for creating, changing, reviewing, reverse-engineering or accepting requirements, PRDs, prototypes, competitor material or existing systems, including any small UI, field, column, tab, dropdown or legacy-HTML change. Supports /ads, /dig, /prd and /proto intent shortcuts where the host routes them. Always invoke…
mvp-build-tickets
Generate implementation task tickets for a Web MVP. Use after AGENTS.md and agentdocs exist when the user needs small Codex-executable task files with required reading, scope, acceptance criteria, implementation plan, verification checks, risks, and done conditions.
mvp-agent-context
Generate Codex-first agent context files for a Web MVP. Use after the PRD and Technical Blueprint exist when the user needs AGENTS.md, REVIEW-CHECKLIST.md, and agentdocs for project brief, requirements, tech stack, page map, data model, build tasks, testing, and reusable prompts.
mvp-workflow
Orchestrate a Codex-first Web MVP planning workflow from rough idea to build-ready artifacts. Use when the user wants to turn an app idea into a structured MVP plan, PRD, technical blueprint, AGENTS.md context, and implementation tickets.
mvp-idea-brief
Create an Idea Brief for a Web MVP. Use when the user has a rough app idea and needs to clarify the target user, problem, current workaround, smallest useful outcome, platform choice, budget, and MVP exclusions before research or PRD work.