agentsclimarketplace

Atomic

Skill Fantasymax/easy-build-skills/plugin/user-research-for-ai-config/skills/user-research-for-ai-config/templates/mode-b/atomic

Use this skill when {B1 一句话定义关键动词}. This includes: {B7 ② 子场景列举,3–5 个,逗号隔开}. Trigger especially when the user mentions {B7 ① 触发原话片段,至少 3 句,原话保留} or files matching {B7 ② 文件名/关键词}. Do NOT use for {B7 ③ 反例 + 通用兜底"general writing requests where the user doesn't want X-specific style"}.From its SKILL.md

Install
npx -y skills add Fantasymax/easy-build-skills --skill atomic

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.1 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

{Skill 标题,人话版,对应 H1}

这是 {用户名} 把"{B1 一句话定义}"这件事抽成的属于自己的 skill。 生成日期 {YYYY-MM-DD} | 版本 V1(快速版)

Overview

{B1 一句话定义原话}

✏️ 写法提示:1–3 句,散文体。第一句直接复用 B1。第二句可补"为什么这件事值得有 skill"。第三句可省略。

Quick Reference

输入输出主要工具
{B2 输入材料关键项}{B1 输出物}{B2 工具}

触发情境(用户原话)

  • "{B7 ① 第 1 句原话}"
  • "{B7 ① 第 2 句原话}"
  • "{B7 ① 第 3 句原话}"

When to use this skill

补充说明(如 description 中已经够具体可省):

  • {B7 ② 子场景说明}

不该触发(务必明示):

  • {B7 ③ 反例 1}
  • {B7 ③ 反例 2}

Workflow

来源:B3 用户的手动 SOP,按时间线排列。每步保留用户原话动词。

Phase 1:{第一阶段名,如"准备"}

  1. {B3 步骤 1 用户原话}

    • 输入:{该步用到的 B2 材料}
    • 注意:{若 B5 行内规矩涉及该步,原话引用}
  2. {B3 步骤 2 用户原话}

    • {若该步是 B4 判断节点 → 标 [需要用户判断]}
    • {判断依据:B4 用户原话规则}

Phase 2:{第二阶段名,如"执行"}

{继续按 B3 步骤填,每步用户原话动词开头}

Phase 3:{第三阶段名,如"收尾"}

{继续}

阶段拆分原则:B3 步骤超过 5 步建议拆 2–3 个 phase;step 命名用动词开头。

Domain Knowledge & Anti-patterns

来源:B5 行内规矩 + B6 禁词禁句。用户原话保留,禁止 AI 改写

必须遵守的"行内规矩"

  • {B5 规矩 1 用户原话}

    • 为什么:{用户解释,原话;如果用户没解释,AI 可补但需用户审核}
  • {B5 规矩 2 用户原话}

    • 为什么:{解释}
  • {B5 规矩 3 用户原话}

    • 为什么:{解释}

Anti-patterns(绝不许这样做)

  • ❌ {B6 ③ 禁词/禁句原话} — 出现这个就是错的
  • ❌ {B5 ① 新人最容易出错的步骤的反向规则}
  • ❌ {B5 ③ 反直觉规则:看似该这样但实际不能}

反例:用户见过的"做得最差"的版本

{B5 ② 用户给的差版本原话片段}

差在哪里:{用户原话或 AI 推断后用户审核}

When to ask the user

来源:B4 判断节点。这些步骤 AI 不该自行决定。

  • {B4 节点 1}:选 A 还是 B 时,依据 {B4 ① 选择内容 + ② 决定依据}。如果不清楚,问用户。
  • {B4 节点 2}:...

Self-check before delivery

来源:B9 用户的交付前自检。如有 ≥ 5 条,单独拆 references/checklist.md

完成前自查:

  • {B9 第 1 条原话}
  • {B9 第 2 条原话}
  • {B9 第 3 条原话}

Reference files

按需读取(Progressive Disclosure):

  • examples/good-1.md — 用户给的好范例原物(B8)
  • examples/bad-with-fix.md — 差范例 + 修订说明(B8)
  • references/glossary.md — 领域术语表(B6)
  • references/checklist.md — 完整自检清单(如 B9 ≥ 5 条时)
  • assets/templates/ — 固定模板(B2 / B6 用户原文)

Notes from author

来源:B10 ① 收尾开放回答。如有用户重要补充,原话保留在此。

{B10 ① 用户原话;如果用户没补充则删掉本段}


元信息(可删除)

  • 生成方式:通过 user-research-for-ai-config skill Mode B 快速版
  • 基于的访谈:共享核心 4 题 + Mode B 10 题
  • 质检状态:description 触发测试 [✓/✗] | dogfood 一次 [✓/✗]
  • 下一次迭代触发条件:dogfood 连续 ≥ 3 次发现"AI 输出不像用户",回到 examples/ 补充

What ships with it: 5 files

8.7 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.