Back to Details

insight-engine

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

insight-engine — the Insight Brief

insight-engine——洞察简报

Fifth station of the pipeline, and the first one that runs after real fieldwork. It takes what real people actually said —interview transcripts, survey answers, support feedback— and converges it into insights that carry their evidence, a customer profile, and opportunities ordered against the business outcome. Generic domain example: a B2B appointment-management app for small clinics, with twelve interviews of patients and administrators already done.
这是流程的第五个环节，也是实地调研完成后首个启动的环节。它会收集真实用户的实际反馈——访谈记录、调查答案、客服反馈——并将其整合为附带证据的洞察、客户画像，以及基于业务成果排序的机会点。通用领域示例：一款面向小型诊所的B2B预约管理应用，已完成12次患者和管理员访谈。

When it is invoked

触发时机

WhenFrom / to
Real fieldwork is done and the raw answers existEntry from real fieldwork — interviews with real people, the work the operator picks up; it is not a skill of this pack. The Diagnosis + Script v2 of 
synthetic-users
(
.claude/skills/synthetic-users/SKILL.md
) is the guide those answers came from
The Insight Brief is writtenExit toward 
ideation-engine
(
.claude/skills/ideation-engine/SKILL.md
) — the solution space starts from the profile this brief validates
With no prior station of the pipelineInvocable on its own, over any set of real answers, with the business outcome as the minimum second input
时机来源/去向
实地调研完成且原始答案已存在输入来自实地调研——与真实用户的访谈，由操作人员负责收集；这不属于本工具包的技能范畴。
synthetic-users
的Diagnosis + Script v2（
.claude/skills/synthetic-users/SKILL.md
）是这些答案收集的指导依据
洞察简报撰写完成输出至
ideation-engine
（
.claude/skills/ideation-engine/SKILL.md
）——解决方案的探索从本简报验证的用户画像开始
无需流程中前置环节可独立调用，适用于任意真实答案数据集，仅需最少第二个输入：业务成果

Rule number one

核心规则

Evidence or silence. Nothing enters the brief without a verbatim behind it, quoted with the identifier of who said it. Where the data does not reach, the brief says "not enough evidence" and names what would have to be researched — it never fills the gap with a plausible sentence. An insight nobody said is an assumption with the authority of research, which is worse than no research.
要么有证据，要么留白。简报中的任何内容都必须附带用户的原始表述，并标注发言者标识。对于数据未覆盖的内容，简报需注明“证据不足”，并说明需要进一步调研的方向——绝不能用看似合理的语句填补空白。没有用户支撑的洞察，是披着研究外衣的假设，比没有研究更糟糕。

The two inputs

两个输入项

The answers (required): one or several sources, in whatever form they came. Converging sources that were collected apart is half the value of this station.
The business outcome (required to prioritize): the objective the product is trying to move. It is asked for with the pressure method of 
.claude/skills/grill/SKILL.md#method
—cited by path, never copied— before scoring anything. Without an outcome there is no prioritization, only a flat list: with no outcome the opportunities are written unordered and the brief says why.
If the answers are simulated or synthetic —because they came from 
synthetic-users
, or because the source says so— the disclaimer travels into the brief: they train the method, they do not decide a roadmap.
答案（必填）：一个或多个来源，无论格式如何。整合分散收集的数据源是本环节的核心价值之一。
业务成果（用于优先级排序，必填）：产品试图达成的目标。在对任何内容评分前，需采用
.claude/skills/grill/SKILL.md#method
中的施压法获取该目标——需通过路径引用，不得复制内容。没有业务成果就无法进行优先级排序，只能得到一个扁平列表：若无业务成果，机会点将无序呈现，且简报需说明原因。
若答案是模拟或合成的——来自
synthetic-users
，或来源明确标注——则需在简报中添加免责声明：这些数据仅用于训练方法，不能用于决定产品路线图。

The convergence, in four steps

整合流程四步骤

1. Clusters. Group by behavior and context, never by demography alone: two people of the same age can have opposite jobs. Each cluster carries a descriptive name, its size as a share of the sample, and its defining traits. Two things are named apart and never quietly dropped: the noise —answers too short, contradictory or empty— counted and excluded out loud, and the outliers, which stay visible: an outlier is sometimes an entire opportunity.
2. Jobs to be done, each labeled by dimension: [FUNCTIONAL] the concrete task, [EMOTIONAL] how the person wants to feel or avoid feeling, [SOCIAL] how they want to be seen. Every job is written as When [context], I want [motivation], so that [expected outcome] and carries at least two literal verbatims with the respondent identifier. The three dimensions are looked for; the one the data does not support is declared missing, never invented.
3. The customer profile. For the prioritized cluster: jobs ordered by importance (frequency × intensity in the data), pains —costs, frictions, risks and negative emotions before, during and after the job— and gains, separating expected from desired from unexpected when the data allows it. Every item with its verbatim. The value map —what the product does about all this— is not written here: it belongs to the solution space and to 
ideation-engine
.
4. Prioritized opportunities. With the outcome as the root, each opportunity is an unmet need expressed by users, never a solution. Each one carries: the statement in the user's own words, the evidence that supports it (which clusters suffer it, which verbatims, which business figure sizes it), and a score against the outcome with the criterion written out and the reasoning visible, not only the number. They close ordered, as the branches that hang from the outcome.
1. 用户分组。需根据行为和场景分组，不能仅依据人口统计特征：年龄相同的两个人可能有完全相反的需求。每个分组需包含描述性名称、在样本中的占比，以及核心特征。以下两项需单独标注，不得随意忽略：干扰项——过短、矛盾或无效的答案，需统计数量并明确排除；异常值——需保留可见：异常值有时本身就是完整的机会点。
2. 用户待办任务，按维度标注：[FUNCTIONAL] 具体任务，[EMOTIONAL] 用户期望的感受或想要避免的感受，[SOCIAL] 用户期望的他人评价。每个任务需以*当[场景]时，我想要[动机]，以便[预期结果]*的格式呈现，并附带至少两条用户原始表述及受访者标识。需覆盖三个维度；若数据不支持某一维度，需明确说明缺失，不得编造。
3. 客户画像。针对优先级最高的用户分组：按重要性排序的任务（数据中的频率×强度）、痛点——任务执行前、中、后的成本、摩擦、风险和负面情绪，以及收益——若数据允许，需区分预期收益、期望收益和意外收益。每个条目都需附带原始表述。价值地图——产品针对上述内容的解决方案——无需在此撰写：它属于解决方案范畴，由
ideation-engine
负责处理。
4. 优先级排序的机会点。以业务成果为核心，每个机会点都是用户表达的未被满足的需求，绝不能是解决方案。每个机会点需包含：用户原话表述、支撑证据（受影响的用户分组、原始表述、相关业务数据规模）、基于业务成果的评分（需明确评分标准及推理过程，不能仅给出分数）。最终按与业务成果的关联度排序。

The Insight Brief

洞察简报

It closes by writing the deliverable as dated research of the loaded product node, in:
context/<YYYY-MM-DD>-insight-brief.md
Dated, never overwritten. The path falls inside 
content: orgs/*/products/*/context/*.md
of 
tree.md
: no new glob and no resolver row are needed.
The brief carries these five sections, in this order, each with its literal heading:
undefined
最终输出为已加载产品节点的时效性研究报告，保存路径为：
context/<YYYY-MM-DD>-insight-brief.md
带有日期标记，永不覆盖。路径位于
tree.md
中的
content: orgs/*/products/*/context/*.md
：无需新增通配符或解析器行。
简报包含以下五个固定顺序的章节，每个章节使用指定标题：
undefined

Clusters and noise

用户分组与干扰项

Jobs to be done

用户待办任务

Customer profile

客户画像

Prioritized opportunities

优先级排序的机会点

What this data does not say

数据未覆盖的内容


- **Clusters and noise**: each cluster with its size and traits, the excluded noise counted, the
  outliers kept and named.
- **Jobs to be done**: every job with its dimension label and its two verbatims.
- **Customer profile**: jobs, pains and gains of the prioritized cluster, each with evidence.
- **Prioritized opportunities**: the branches hanging from the outcome, with score and reasoning.
- **What this data does not say**: the limits of the sample, the biases the collection could carry,
  and what would have to be researched before building. This section is written every time, also
  when the findings look conclusive — especially then.

- **用户分组与干扰项**：每个分组的规模和特征，已排除的干扰项数量，保留并标注的异常值。
- **用户待办任务**：每个任务的维度标签及两条原始表述。
- **客户画像**：优先级最高分组的任务、痛点和收益，每项均附带证据。
- **优先级排序的机会点**：以业务成果为核心的分支列表，包含评分及推理过程。
- **数据未覆盖的内容**：样本的局限性、数据收集可能存在的偏差，以及开发前需进一步调研的内容。无论研究结果看似多么确凿，此章节每次都必须撰写——尤其在结果看似确凿时更需如此。

What this deliverable does not claim

本成果的局限性

The Insight Brief does not say what to build: it says what is worth attacking and with what evidence. Turning an opportunity into ideas, and ideas into a value proposition, is the job of 
ideation-engine
. Deciding the scope and writing the strategic layer of the product is the job of 
prd
(
.claude/skills/prd/SKILL.md
).
洞察简报不会说明要构建什么：它只会说明哪些内容值得投入，以及对应的证据。将机会点转化为创意，再将创意转化为价值主张，是
ideation-engine
的工作。决定产品范围并撰写产品战略层内容，是
prd
（
.claude/skills/prd/SKILL.md
）的工作。

Standalone use

独立使用方式

Without a brain, the deliverable is written to a file in the current folder: the process is the same, but nothing is filed into a node, a backlog or a resolver, and no session scan brings it back later. Inside AI First OS the destinations named above are the real ones and this section does not apply.
The rest of the pack installs one skill at a time. Look at 
.claude/skills/
first and offer only the ones that are not there — inside AI First OS they are all installed already, so nothing here gets offered:
  • grill
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill grill
  • ideation-engine
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill ideation-engine
  • prd
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill prd
  • synthetic-users
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill synthetic-users
The whole pack at once: 
npx skills add pedroromeroluna/ai-first-product-skills
.
The system these skills come from is AI First OS — 
github.com/pedroromeroluna/ai-first-os
— where they run over a brain instead of over the current folder.
若未接入AI First OS，成果将保存至当前文件夹的文件中：流程与上述一致，但不会归档至节点、待办事项或解析器，后续也无法通过会话扫描找回。在AI First OS中，上述目标路径为默认路径，本章节不适用。
工具包中的其余技能需逐个安装。请先查看
.claude/skills/
，仅安装未存在的技能——在AI First OS中所有技能已预装，因此无需安装以下技能：
  • grill
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill grill
  • ideation-engine
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill ideation-engine
  • prd
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill prd
  • synthetic-users
    — 
    npx skills add pedroromeroluna/ai-first-product-skills --skill synthetic-users
一键安装整个工具包：
npx skills add pedroromeroluna/ai-first-product-skills
。
这些技能源自AI First OS——
github.com/pedroromeroluna/ai-first-os
——在该系统中，技能运行于AI大脑而非当前文件夹。