company-research

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Company Research

企业调研

Discover and deeply research companies to sell to. Uses Browserbase Search API for discovery and a Plan→Research→Synthesize pattern for deep enrichment — outputting a scored research report and CSV.

Required:

BROWSERBASE_API_KEY

env var and

browse

CLI installed.

First-run setup: On the first run you'll be prompted to approve

browse cloud fetch

browse cloud search

cat

mkdir

sed

, etc. Select "Yes, and don't ask again for: browse cloud fetch:*" (or equivalent) for each to auto-approve for the session. To permanently approve, add these to your

~/.claude/settings.json

under

permissions.allow

json

"Bash(browse:*)", "Bash(bunx:*)", "Bash(bun:*)", "Bash(node:*)",
"Bash(cat:*)", "Bash(mkdir:*)", "Bash(sed:*)", "Bash(head:*)", "Bash(tr:*)", "Bash(rm:*)"

Path rules: Always use the full literal path in all Bash commands — NOT

$HOME

(both trigger "shell expansion syntax" approval prompts). Resolve the home directory once and use it everywhere. When constructing subagent prompts, replace

{SKILL_DIR}

with the full literal path.

Output directory: All research output goes to

~/Desktop/{company_slug}_research_{YYYY-MM-DD}/

. This directory contains one

.md

file per researched company plus a final

.csv

. The user gets both the scored spreadsheet and the full research files on their Desktop.

CRITICAL — Tool restrictions (applies to main agent AND all subagents):

All web searches: use
```
browse cloud search
```
. NEVER use WebSearch.
All page content extraction: use
```
node {SKILL_DIR}/scripts/extract_page.mjs "<url>"
```
. This script fetches via
```
browse cloud fetch --output
```
, parses title + meta tags + visible body text, and automatically falls back to
```
browse get markdown
```
when fetch fails or returns thin JS-rendered content. NEVER hand-roll a
```
browse cloud fetch | sed
```
pipeline — it strips meta tags and doesn't parse the stdout JSON envelope. NEVER use WebFetch.
All research output: subagents write one markdown file per company to
```
{OUTPUT_DIR}/{company-slug}.md
```
using bash heredoc. NEVER use the Write tool or
```
python3 -c
```
. See
```
references/example-research.md
```
for the file format.
Report + CSV compilation: use
```
node {SKILL_DIR}/scripts/compile_report.mjs {OUTPUT_DIR} --open
```
— generates HTML report and CSV in one step, opens overview in browser.

URL deduplication: use

node {SKILL_DIR}/scripts/list_urls.mjs /tmp

after discovery.

Subagents must use ONLY the Bash tool. No other tools allowed.
Main agent NEVER reads raw discovery JSON batch files. Use
```
list_urls.mjs
```
for dedup.

CRITICAL — Anti-hallucination rules (applies to main agent AND all subagents):

NEVER infer
```
product_description
```
,
```
industry
```
, or
```
target_audience
```
from a site's fonts, framework (Framer/Next.js/React), design system, or typography. These are cosmetic and say nothing about what the company sells.
NEVER let the user's own ICP leak into a target's description. If you don't know what the target does, write
```
Unknown
```
— do not pattern-match them onto the ICP.
```
product_description
```
MUST quote or paraphrase a specific phrase from
```
extract_page.mjs
```
output (TITLE, META_DESCRIPTION, OG_DESCRIPTION, HEADINGS, or BODY). If none of those fields yield a recognizable product statement, write
```
Unknown — homepage content not accessible
```
.

product_description

Unknown

, cap

icp_fit_score

at 3 and set

icp_fit_reasoning

Insufficient evidence — homepage returned no readable content

CRITICAL — Minimize permission prompts:

Subagents MUST batch ALL file writes into a SINGLE Bash call using chained heredocs. One Bash call = one permission prompt.
Batch ALL searches and ALL fetches into single Bash calls using
```
&&
```
chaining.

发掘并深度研究可销售的目标企业。借助Browserbase Search API进行企业发掘，采用「规划→调研→合成」模式进行深度信息补充，最终输出带评分的研究报告及CSV文件。

必要条件：需配置

BROWSERBASE_API_KEY

环境变量，且已安装

browse

CLI工具。

首次运行设置：首次运行时，系统会提示您批准

browse cloud fetch

、

browse cloud search

、

cat

、

mkdir

、

sed

等命令。请选择「是，且不再针对以下命令询问：browse cloud fetch:*」（或类似选项），以便在本次会话中自动批准。若要永久批准，请将这些命令添加到

~/.claude/settings.json

的

permissions.allow

字段中：

json

"Bash(browse:*)", "Bash(bunx:*)", "Bash(bun:*)", "Bash(node:*)",
"Bash(cat:*)", "Bash(mkdir:*)", "Bash(sed:*)", "Bash(head:*)", "Bash(tr:*)", "Bash(rm:*)"

路径规则：在所有Bash命令中必须使用完整的字面路径——禁止使用

或

$HOME

（两者会触发「shell扩展语法」批准提示）。只需解析一次主目录，之后统一使用该路径。构建子Agent提示词时，将

{SKILL_DIR}

替换为完整的字面路径。

输出目录：所有调研输出文件将保存至

~/Desktop/{company_slug}_research_{YYYY-MM-DD}/

。该目录包含每个被调研企业对应的

.md

文件，以及最终的

.csv

文件。用户可在桌面获取带评分的电子表格及完整调研文件。

重要提示——工具限制（适用于主Agent及所有子Agent）：

所有网页搜索：使用
```
browse cloud search
```
。禁止使用WebSearch。
所有页面内容提取：使用
```
node {SKILL_DIR}/scripts/extract_page.mjs "<url>"
```
。该脚本通过
```
browse cloud fetch --output
```
获取内容，解析标题、元标签及可见正文文本，当获取失败或返回内容为轻量JS渲染时，会自动回退至
```
browse get markdown
```
。禁止手动编写
```
browse cloud fetch | sed
```
管道——该方式会剥离元标签且无法解析标准输出的JSON包。禁止使用WebFetch。
所有调研输出：子Agent需使用bash heredoc将每个企业的调研内容写入单独的markdown文件，保存至
```
{OUTPUT_DIR}/{company-slug}.md
```
。禁止使用Write工具或
```
python3 -c
```
。文件格式可参考
```
references/example-research.md
```
。
报告与CSV编译：使用
```
node {SKILL_DIR}/scripts/compile_report.mjs {OUTPUT_DIR} --open
```
——一步生成HTML报告及CSV文件，并在浏览器中打开概览页面。

URL去重：在发掘完成后使用

node {SKILL_DIR}/scripts/list_urls.mjs /tmp

。

子Agent只能使用Bash工具，禁止使用其他工具。
主Agent禁止读取原始的发掘JSON批量文件，请使用
```
list_urls.mjs
```
进行去重。

重要提示——防幻觉规则（适用于主Agent及所有子Agent）：

禁止从网站的字体、框架（Framer/Next.js/React）、设计系统或排版推断
```
product_description
```
、
```
industry
```
或
```
target_audience
```
。这些属于外观层面，无法反映企业的业务内容。
禁止将用户自身的ICP内容混入目标企业的描述中。若不清楚目标企业的业务，请填写
```
Unknown
```
——切勿将其强行匹配到ICP中。
```
product_description
```
必须引用或改写
```
extract_page.mjs
```
输出中的特定内容（TITLE、META_DESCRIPTION、OG_DESCRIPTION、HEADINGS或BODY）。若这些字段均未提供可识别的产品说明，请填写
```
Unknown — homepage content not accessible
```
。

若

product_description

为

Unknown

，则

icp_fit_score

最高只能设为3，且

icp_fit_reasoning

需设置为

Insufficient evidence — homepage returned no readable content

。

重要提示——最小化权限提示：

子Agent必须使用链式heredoc将所有文件写入操作批量处理为单次Bash调用。一次Bash调用对应一次权限提示。
使用
```
&&
```
链式调用将所有搜索及获取操作批量处理为单次Bash调用。

Pipeline Overview

流程概览

Follow these 5 steps in order. Do not skip steps or reorder.

Company Research — Deeply understand the user's company, product, and who they sell to
Depth Mode Selection — Choose research depth based on how many targets they want
Discovery — Find target companies using diverse search queries
Deep Research & Scoring — Research each company, score ICP fit
Report & CSV — Present findings, compile scored CSV

请按以下5个步骤依次执行，请勿跳过或调整顺序。

企业调研——深度了解用户的企业、产品及销售对象
深度模式选择——根据目标企业数量选择调研深度
企业发掘——使用多样化搜索查询寻找目标企业
深度调研与评分——调研每个企业，对ICP匹配度打分
报告与CSV生成——展示调研结果，编译带评分的CSV文件

Step 0: Setup Output Directory

步骤0：设置输出目录

Before starting, create the output directory on the user's Desktop:

bash

OUTPUT_DIR=~/Desktop/{company_slug}_research_{YYYY-MM-DD}
mkdir -p "$OUTPUT_DIR"

Replace

{company_slug}

with the user's company name (lowercase, hyphenated) and

{YYYY-MM-DD}

with today's date. Pass

{OUTPUT_DIR}

(as a full literal path, not with

) to all subagent prompts so they write research files there.

Also clean up discovery batch files from prior runs:

bash

rm -f /tmp/company_discovery_batch_*.json

开始前，请在用户桌面创建输出目录：

bash

OUTPUT_DIR=~/Desktop/{company_slug}_research_{YYYY-MM-DD}
mkdir -p "$OUTPUT_DIR"

将

{company_slug}

替换为用户企业名称（小写，连字符分隔），

{YYYY-MM-DD}

替换为当前日期。将

{OUTPUT_DIR}

（完整字面路径，请勿使用

）传递给所有子Agent的提示词，以便它们将调研文件写入该目录。

同时清理之前运行产生的发掘批量文件：

bash

rm -f /tmp/company_discovery_batch_*.json

Step 1: Deep Company Research

步骤1：深度企业调研

This is the most important step. The quality of everything downstream depends on deeply understanding the user's company.

Ask the user for their company name or URL
Check for an existing profile:
- List files in
```
{SKILL_DIR}/profiles/
```
  (ignore
```
example.json
```
  )
- If a matching profile exists → load it, present to user: "I have your profile from {researched_at}. Still accurate?" If yes → skip to Step 2.
- If no profile exists → proceed with deep research below.
Run a full deep research on the user's company using the Plan→Research→Synthesize pattern. See
```
references/research-patterns.md
```
for sub-question templates and research methodology.
Key research steps:
- Search:
```
browse cloud search "{company name}" --num-results 10
```
- Fetch homepage:
```
node {SKILL_DIR}/scripts/extract_page.mjs "{company website}"
```
- Discover site pages via sitemap (do NOT hardcode paths like
```
/about
```
  or
```
/customers
```
  ):
  1. ```
  browse cloud fetch --allow-redirects "{company website}/sitemap.xml"
```
  — sitemap is small, raw
```
  browse cloud fetch
```
  is fine
2. Scan for URLs with keywords:
```
  customer
```
  ,
```
  case-stud
```
  ,
```
  pricing
```
  ,
```
  about
```
  ,
```
  use-case
```
  ,
```
  industry
```
  ,
```
  solution
```
3. Optionally also fetch
```
  /llms.txt
```
  for page descriptions
4. Pick 3-5 most relevant URLs and extract with
```
  extract_page.mjs
```
  (NOT raw
```
  browse cloud fetch
```
  )
- Search for external context and competitors
- Accumulate findings with confidence levels
Synthesize into a profile: Company, Product, Existing Customers, Competitors, Use Cases. Do NOT include ICP or sub-verticals — those are per-run decisions.
Present the profile to the user for confirmation. Do not proceed until confirmed.

Save the confirmed profile to

{SKILL_DIR}/profiles/{company-slug}.json

Ask clarifying questions using
```
AskUserQuestion
```
with checkboxes:
- "Which segments are you targeting?" with options derived from the company research
- "Company stage?" — Startups, Mid-market, Enterprise, All
- "How many companies / depth?" — Quick (~100), Deep (~50), Deeper (~25)
- This is the ONLY user interaction. After this, execute silently until results are ready.

这是最重要的步骤，后续所有环节的质量都取决于对用户企业的深度理解。

向用户询问其企业名称或网址
检查是否存在现有档案：
- 列出
```
{SKILL_DIR}/profiles/
```
  中的文件（忽略
```
example.json
```
  ）
- 若存在匹配的档案→加载该档案并展示给用户：「我有您于{researched_at}创建的档案，是否仍然准确？」若用户确认是→跳至步骤2。
- 若不存在匹配的档案→继续进行下方的深度调研。
采用「规划→调研→合成」模式对用户企业进行全面深度调研。子问题模板及调研方法可参考
```
references/research-patterns.md
```
。
关键调研步骤：
- 搜索：
```
browse cloud search "{company name}" --num-results 10
```
- 获取主页内容：
```
node {SKILL_DIR}/scripts/extract_page.mjs "{company website}"
```
- 通过站点地图发掘页面（请勿硬编码
```
/about
```
  或
```
/customers
```
  等路径）：
  1. ```
  browse cloud fetch --allow-redirects "{company website}/sitemap.xml"
```
  ——站点地图内容较少，直接使用
```
  browse cloud fetch
```
  即可
2. 扫描包含以下关键词的URL：
```
  customer
```
  、
```
  case-stud
```
  、
```
  pricing
```
  、
```
  about
```
  、
```
  use-case
```
  、
```
  industry
```
  、
```
  solution
```
3. 可选：获取
```
  /llms.txt
```
  以获取页面描述
4. 选择3-5个最相关的URL，使用
```
  extract_page.mjs
```
  提取内容（请勿使用原始
```
  browse cloud fetch
```
  ）
- 搜索外部背景信息及竞争对手
- 收集调研结果并标注置信度
合成档案：内容包括企业、产品、现有客户、竞争对手、使用场景。请勿包含ICP或细分垂直领域——这些属于每次运行时的决策内容。
将档案展示给用户确认，得到确认后再继续。

将确认后的档案保存至

{SKILL_DIR}/profiles/{company-slug}.json

使用
AskUserQuestion
工具并提供复选框，询问澄清问题：
- 「您的目标细分领域是？」选项基于企业调研结果生成
- 「企业阶段？」——初创企业、中型市场、企业级、全部
- 「目标企业数量/调研深度？」——快速（约100家）、深度（约50家）、更深度（约25家）
- 这是唯一需要用户交互的环节。完成后将静默执行直至生成结果。

Step 2: Depth Mode Selection

步骤2：深度模式选择

Mode	Research per company	Best for
`quick`	Homepage + 1-2 searches	~100 companies, broad scan
`deep`	2-3 sub-questions, 5-8 tool calls	~50 companies, solid research
`deeper`	4-5 sub-questions, 10-15 tool calls	~25 companies, full intelligence

模式	单企业调研内容	适用场景
`quick`	主页+1-2次搜索	约100家企业，广泛扫描
`deep`	2-3个子问题，5-8次工具调用	约50家企业，扎实调研
`deeper`	4-5个子问题，10-15次工具调用	约25家企业，全面情报收集

Step 3: Discovery

步骤3：企业发掘

Formula:

ceil(requested_companies / 35)

search queries needed. Over-discover by ~2-3x because filtering typically drops 50-70%.

Generate search queries with these patterns:

Industry + company stage + geography ("fintech startups series A Bay Area")
Technology stack + use case ("companies using Selenium for web scraping")
Competitor adjacency ("alternatives to {known company in ICP}")
Buyer persona + pain point ("engineering teams struggling with browser automation")

Process:

Launch ALL discovery subagents at once (up to ~6 per message). Each runs its queries in a SINGLE Bash call:
bash
```
browse cloud search "{query}" --num-results 25 --output /tmp/company_discovery_batch_{N}.json
```

After all waves complete, deduplicate:

node {SKILL_DIR}/scripts/list_urls.mjs /tmp

Filter the URL list — remove:
- Blog posts, news articles (globenewswire.com, techcrunch.com, etc.)
- Directories/aggregators (tracxn.com, crunchbase.com, g2.com)
- The user's own competitors and existing customers (from profile) Keep only company homepages.

See

references/workflow.md

for subagent prompt templates and wave management.

公式：需要

ceil(requested_companies / 35)

个搜索查询。需超额发掘2-3倍，因为过滤通常会剔除50-70%的结果。

使用以下模式生成搜索查询：

行业+企业阶段+地域（例如："fintech startups series A Bay Area"）
技术栈+使用场景（例如："companies using Selenium for web scraping"）
竞品关联（例如："alternatives to {known company in ICP}"）
买家角色+痛点（例如："engineering teams struggling with browser automation"）

流程：

同时启动所有发掘子Agent（每条消息最多约6个）。每个子Agent在单次Bash调用中运行其查询：
bash
```
browse cloud search "{query}" --num-results 25 --output /tmp/company_discovery_batch_{N}.json
```

所有批次完成后，进行去重：

node {SKILL_DIR}/scripts/list_urls.mjs /tmp

过滤URL列表——移除：
- 博客文章、新闻稿（如globenewswire.com、techcrunch.com等）
- 目录/聚合平台（如tracxn.com、crunchbase.com、g2.com）
- 用户自身的竞争对手及现有客户（来自档案）仅保留企业主页。

子Agent提示词模板及批次管理可参考

references/workflow.md

。

Step 4: Deep Research & Scoring

步骤4：深度调研与评分

Launch subagents to research companies in parallel. See

references/workflow.md

for the enrichment subagent prompt template. See

references/research-patterns.md

for the full research methodology.

Process:

Split filtered URLs into groups per subagent (quick: ~10, deep: ~5, deeper: ~2-3)
Launch ALL enrichment subagents at once (up to ~6 per message)
Each subagent uses ONLY Bash — for each company:

Phase A — Plan (skip in quick mode): Decompose into 2-5 sub-questions based on ICP and enrichment fields.

Phase B — Research Loop: Search and fetch pages, extract findings. Respect step budget (quick: 2-3, deep: 5-8, deeper: 10-15).

Phase C — Synthesize: Score ICP fit 1-10 with evidence. Fill enrichment fields from findings.
Subagents write ALL markdown files in a SINGLE Bash call using chained heredocs to
```
{OUTPUT_DIR}/
```
After ALL subagents complete, proceed to Step 5

Critical: Include the confirmed ICP description verbatim in every subagent prompt. Pass the full literal

{OUTPUT_DIR}

path to every subagent.

并行启动子Agent进行企业调研。信息补充子Agent提示词模板可参考

references/workflow.md

，完整调研方法可参考

references/research-patterns.md

。

流程：

将过滤后的URL按子Agent分组（快速模式：约10家/组，深度模式：约5家/组，更深度模式：约2-3家/组）
同时启动所有信息补充子Agent（每条消息最多约6个）
每个子Agent仅使用Bash工具——针对每个企业：

阶段A——规划（快速模式跳过）：根据ICP及信息补充字段分解为2-5个子问题。

阶段B——调研循环：搜索并获取页面内容，提取调研结果。遵守步骤预算（快速模式：2-3步，深度模式：5-8步，更深度模式：10-15步）。

阶段C——合成：对ICP匹配度打1-10分并提供依据。根据调研结果填写信息补充字段。
子Agent使用链式heredoc将所有markdown文件通过单次Bash调用写入
```
{OUTPUT_DIR}/
```
所有子Agent完成后，进入步骤5

重要提示：在每个子Agent的提示词中逐字包含已确认的ICP描述。将完整的字面

{OUTPUT_DIR}

路径传递给每个子Agent。

Step 5: Report & CSV

步骤5：报告与CSV生成

Generate HTML report + CSV (opens overview in browser automatically):
bash
```
node {SKILL_DIR}/scripts/compile_report.mjs {OUTPUT_DIR} --open
```
This generates:
- ```
{OUTPUT_DIR}/index.html
```
  — overview page with scored table (opens in browser)
- ```
{OUTPUT_DIR}/companies/*.html
```
  — individual company pages (linked from overview)
- ```
{OUTPUT_DIR}/results.csv
```
  — scored spreadsheet for import into sheets/CRM
Present a summary in chat too:

undefined

生成HTML报告+CSV文件（自动在浏览器中打开概览页面）：
bash
```
node {SKILL_DIR}/scripts/compile_report.mjs {OUTPUT_DIR} --open
```
该命令将生成：
- ```
{OUTPUT_DIR}/index.html
```
  ——带评分表格的概览页面（将在浏览器中打开）
- ```
{OUTPUT_DIR}/companies/*.html
```
  ——单个企业的详情页面（从概览页面链接）
- ```
{OUTPUT_DIR}/results.csv
```
  ——带评分的电子表格，可导入表格工具或CRM系统
同时在聊天窗口展示总结内容：

undefined

Company Research Complete

企业调研完成

Total companies researched: {count}
Depth mode: {mode}
Score distribution:
- Strong fit (8-10): {count}
- Partial fit (5-7): {count}
- Weak fit (1-4): {count}
Report opened in browser: ~/Desktop/{company_slug}research{date}/index.html


3. Show the **top companies** sorted by ICP score in a table:

Company	Score	Product	Industry	Fit Reasoning
Acme	9	AI inventory management	E-commerce SaaS	Series A, uses Selenium, expanding to EU


4. For the top 3-5 companies, show a brief research summary — key findings, why they're a good fit, and what specific angle to approach them with.

Offer to dig deeper into specific companies, adjust scoring criteria, or re-run discovery with different queries.

调研企业总数：{count}
深度模式：{mode}
评分分布：
- 高匹配度（8-10分）：{count}
- 部分匹配度（5-7分）：{count}
- 低匹配度（1-4分）：{count}
报告已在浏览器中打开：~/Desktop/{company_slug}research{date}/index.html


3. 展示按ICP评分排序的**top企业**表格：

企业名称	评分	产品	行业	匹配理由
Acme	9	AI库存管理	电商SaaS	A轮融资，使用Selenium，正拓展欧洲市场


4. 针对排名前3-5的企业，展示简短的调研总结——关键发现、匹配原因及具体沟通切入点。

主动提出可为特定企业进行更深层次调研、调整评分标准或使用不同查询重新进行企业发掘。