nimble-web-expert

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Nimble Web Expert

Nimble Web 专家

Web extraction, search, and URL discovery using the Nimble CLI. Returns clean structured data from any website.

User request: $ARGUMENTS

借助Nimble CLI实现网页提取、搜索与URL发现。从任意网站返回整洁的结构化数据。

用户请求：$ARGUMENTS

Core principles

核心原则

Route by intent first. Named site/domain → check for pre-built agent first (announce it out loud). Direct URL →
```
nimble extract
```
. Research/topic →
```
nimble search
```
. Discover/crawl URLs →
```
nimble map
```
or
```
nimble crawl
```
.
One command → present results → done. Run once with
```
--transform "data.parsing"
```
for agents. Show the data immediately as a table. Do NOT experiment, loop, or write Python to parse output.
Multiple inputs → always parallel. 2+ URLs/keywords/ASINs →
```
&
```
+
```
wait
```
. 6–20 →
```
xargs -P
```
. 20+ → Python asyncio script. See
```
references/batch-patterns.md
```
.
Escalate render tiers silently. Tier 1 → 2 → 3 → … without asking. Surface a decision only when all tiers fail and investigation tools are needed.
Never answer from training data. Live prices, current news, today's listings → always fetch via Nimble. If unavailable, say so.
AskUserQuestion at every meaningful choice. Header ≤12 chars, 2–4 options, label 1–5 words, recommended option first. Never present choices as numbered prose.
Save all outputs to
.nimble/
. Never leave extraction results in memory only.
If bash is denied, stop immediately. Show the command as text and wait. Never retry with
```
dangerouslyDisableSandbox
```
.

优先按意图路由。指定网站/域名→先检查是否有预构建agent（大声告知）。直接URL→
```
nimble extract
```
。研究/主题→
```
nimble search
```
。发现/爬取URL→
```
nimble map
```
或
```
nimble crawl
```
。
一条命令→呈现结果→完成。使用
```
--transform "data.parsing"
```
运行agent。立即以表格形式展示数据。请勿尝试实验、循环或编写Python解析输出。
多输入→始终并行处理。2个及以上URL/关键词/ASIN→
```
&
```
+
```
wait
```
。6–20个→
```
xargs -P
```
。20个以上→Python asyncio脚本。详见
```
references/batch-patterns.md
```
。
静默升级渲染层级。从层级1→2→3→…无需询问。仅当所有层级都失败且需要调查工具时，才告知决策。
绝不从训练数据中作答。实时价格、最新新闻、当日列表→始终通过Nimble获取。若无法获取，如实告知。
在每个关键选择点使用AskUserQuestion。标题≤12字符，2–4个选项，标签1–5个词，推荐选项排在首位。切勿以编号段落形式呈现选项。
将所有输出保存至
.nimble/
。切勿仅将提取结果留在内存中。
若bash被禁用，立即停止。以文本形式展示命令并等待。切勿使用
```
dangerouslyDisableSandbox
```
重试。

Skill ecosystem

技能生态系统

Skill	Best for	Key commands
nimble-web-expert (this skill)	Real-time data — fetch any URL, search, map, crawl, run existing agents	`extract` , `search` , `map` , `crawl` , `agent run`
nimble-agent-builder	Build reusable agents — create, refine, publish named extraction templates	CLI: `generate` , `get-generation` , `publish`

Hand off to nimble-agent-builder only when all of these are true: the user has signalled a recurring/scheduled need, the pattern is repetitive (same site, same fields), and they've seen and approved the results. Don't ask after every extract — only when language clearly signals a recurring workflow ("I want to do this every day", "build me a pipeline", "make this reusable").

For agent refinement: "Agent updates are handled by nimble-agent-builder — it can refine the existing agent without rebuilding from scratch."

技能	适用场景	核心命令
nimble-web-expert（本技能）	实时数据——获取任意URL、搜索、映射、爬取、运行现有agent	`extract` , `search` , `map` , `crawl` , `agent run`
nimble-agent-builder	构建可复用agent——创建、优化、发布命名提取模板	CLI: `generate` , `get-generation` , `publish`

仅当以下所有条件满足时，才转交至nimble-agent-builder：用户表示有重复/定时需求，模式具有重复性（同一网站、同一字段），且用户已查看并认可结果。不要在每次提取后都询问——仅当语言明确表明需要重复工作流时才询问（如“我想每天做这个”、“帮我构建一个管道”、“让这个可复用”）。

关于agent优化："Agent更新由nimble-agent-builder处理——它无需从头重建即可优化现有agent。"

Interactive UX

交互式用户体验

Use
```
AskUserQuestion
```
at every meaningful choice — never guess, never ask in prose.
Ambiguous request (no URL, vague topic): ask before running — "What would you like to do?" → Search / Fetch URL / Discover URLs / Call API
Before running a search (if task maps to a specific focus mode): offer focus mode — General / News / Coding / Shopping / Academic / Social
After all tiers fail: check investigation tools (
```
which browser-use
```
,
```
python3 -c "from playwright.sync_api..."
```
) and ask whether to investigate with browser-use, Playwright, or skip.
After presenting results, always close with: "Were these results what you needed?" →
```
Looks great!
```
/
```
Mostly good
```
/
```
Not quite
```
/
```
Skip feedback
```

在每个关键选择点使用
```
AskUserQuestion
```
——切勿猜测，切勿以散文形式询问。
模糊请求（无URL、主题模糊）：运行前询问——“您想执行什么操作？”→搜索 / 获取URL / 发现URL / 调用API
运行搜索前（若任务对应特定聚焦模式）：提供聚焦模式选项——通用 / 新闻 / 编程 / 购物 / 学术 / 社交
所有层级失败后：检查调查工具（
```
which browser-use
```
,
```
python3 -c "from playwright.sync_api..."
```
）并询问是否使用browser-use、Playwright进行调查，或跳过。
呈现结果后，始终以以下问题收尾：“这些结果符合您的需求吗？”→
```
看起来不错！
```
/
```
大体符合
```
/
```
不太符合
```
/
```
跳过反馈
```

Prerequisites

前置条件

Quick check:

bash

nimble --version && echo "${NIMBLE_API_KEY:+API key: set}"

If CLI version and

API key: set

both print → proceed to Step 0.

If anything is missing, load

rules/setup.md

for one-time setup instructions (CLI install, API key, Docs MCP).

If bash is denied: Stop. Show the command as text. Do not substitute WebFetch for Nimble tasks.

快速检查：

bash

nimble --version && echo "${NIMBLE_API_KEY:+API key: set}"

若CLI版本和

API key: set

均正常输出→继续至步骤0。

若有缺失，加载

rules/setup.md

获取一次性设置说明（CLI安装、API密钥、Docs MCP）。

若bash被禁用：停止操作。以文本形式展示命令。切勿用WebFetch替代Nimble任务。

Analyze & Route

分析与路由

User signal	Command	Notes
Names a specific site or domain	`nimble agent` → `nimble extract` if no agent	Always check for agent first — announce it
Provides a direct URL	`nimble extract`	Skip agent check
Research, topic, or vertical query	`nimble search`	Use focus modes for news, jobs, shopping, etc.
"Find URLs / sitemap / all pages"	`nimble map`	Returns URL list + metadata
"Crawl / archive a whole section"	`nimble crawl`	Async bulk extraction

用户信号	命令	备注
指定特定网站或域名	`nimble agent` → 若无agent则使用 `nimble extract`	始终先检查agent——告知用户
提供直接URL	`nimble extract`	跳过agent检查
研究、主题或垂直查询	`nimble search`	针对新闻、职位、购物等场景使用聚焦模式
"查找URL / 站点地图 / 所有页面"	`nimble map`	返回URL列表+元数据
"爬取 / 归档整个板块"	`nimble crawl`	异步批量提取

Step 0 — Agent check (when a domain is named)

步骤0 — Agent检查（当指定域名时）

Pre-built agents return clean structured data with zero selector work. Always check first.

Always verbalize — never silently:

Announce: "Let me check if there's a pre-built Nimble agent for [site]..."
Report: "Found
<agent_name>
— using it now." or "No pre-built agent — falling back to extraction."

Lookup order:

~/.claude/skills/nimble-web-expert/learned/examples.json

→

agents[]

array

```
references/nimble-agents/SKILL.md
```
→ baked-in table (50+ sites)

nimble agent list --limit 100 --search "<domain or vertical>"

→ show table, confirm with user

No match → proceed to extract/search

Run with
--transform "data.parsing"
— always:

bash

nimble --transform "data.parsing" agent run --agent <name> --params '{"keyword": "..."}'

Do NOT run without

--transform "data.parsing"

and then parse raw output. The raw response contains

html

(useless),

headers

, and

parsing

(what you want). The transform flag extracts

parsing

in one shot.

For the full agent list (50+ sites), see

references/nimble-agents/SKILL.md

⚠️

google_search

is for SEO/SERP rank analysis only — not general information retrieval. For finding information, use

nimble search

预构建agent可返回整洁的结构化数据，无需选择器配置。始终先检查。

务必告知用户——切勿静默操作：

告知："让我检查是否有针对[网站]的预构建Nimble agent..."
反馈："找到
<agent_name>
——正在使用。" 或 "无预构建agent——将回退至提取模式。"

查找顺序：

~/.claude/skills/nimble-web-expert/learned/examples.json

→

agents[]

数组

```
references/nimble-agents/SKILL.md
```
→ 内置表格（50+个网站）

nimble agent list --limit 100 --search "<domain or vertical>"

→ 展示表格，确认用户

无匹配项→继续提取/搜索

始终使用
--transform "data.parsing"
运行：

bash

nimble --transform "data.parsing" agent run --agent <name> --params '{"keyword": "..."}'

切勿不使用

--transform "data.parsing"

就运行然后解析原始输出。原始响应包含

html

（无用）、

headers

和

parsing

（所需内容）。转换标志可一次性提取

parsing

内容。

完整agent列表（50+个网站）详见

references/nimble-agents/SKILL.md

。

⚠️

google_search

仅用于SEO/SERP排名分析——不用于通用信息检索。查找信息请使用

nimble search

。

Workflow

工作流

Situation	Command	Reference
Site/domain → check agent first	`nimble agent list` → `nimble agent run`	`references/nimble-agents/SKILL.md`
Direct URL	`nimble extract`	`references/nimble-extract/SKILL.md`
Search the live web	`nimble search`	`references/nimble-search/SKILL.md`
Discover URLs on a site	`nimble map`	`references/nimble-map/SKILL.md`
Bulk crawl a section	`nimble crawl run`	`references/nimble-crawl/SKILL.md`
Batch agents (up to 1,000)	`nimble agent run-batch`	`references/nimble-agents/SKILL.md`
Batch extract (up to 1,000)	`nimble extract-batch`	`references/nimble-extract/SKILL.md`
Poll tasks / batches / results	`nimble tasks` / `nimble batches`	`references/nimble-tasks/SKILL.md`
Unknown selectors or XHR path	browser-use or Playwright investigation	`references/nimble-extract/browser-investigation.md`
Proven site patterns	copy a recipe	`references/recipes.md`
2+ inputs	parallel bash `&` + `wait` or generated script	`references/batch-patterns.md`

For the full extract waterfall (tiers, flags, browser actions, network capture), see

references/nimble-extract/SKILL.md

场景	命令	参考文档
网站/域名→先检查agent	`nimble agent list` → `nimble agent run`	`references/nimble-agents/SKILL.md`
直接URL	`nimble extract`	`references/nimble-extract/SKILL.md`
实时网页搜索	`nimble search`	`references/nimble-search/SKILL.md`
发现网站上的URL	`nimble map`	`references/nimble-map/SKILL.md`
批量爬取板块	`nimble crawl run`	`references/nimble-crawl/SKILL.md`
批量运行agent（最多1000个）	`nimble agent run-batch`	`references/nimble-agents/SKILL.md`
批量提取（最多1000个）	`nimble extract-batch`	`references/nimble-extract/SKILL.md`
轮询任务/批量/结果	`nimble tasks` / `nimble batches`	`references/nimble-tasks/SKILL.md`
未知选择器或XHR路径	使用browser-use或Playwright调查	`references/nimble-extract/browser-investigation.md`
已验证的网站模式	复制示例脚本	`references/recipes.md`
2个及以上输入	并行bash `&` + `wait` 或生成脚本	`references/batch-patterns.md`

完整提取流程（层级、标志、浏览器操作、网络捕获）详见

references/nimble-extract/SKILL.md

。

Response shapes

响应格式

Command	Output
`nimble agent`	Structured data in `data.parsing` — array (SERP/list) or dict (PDP/product)
`nimble extract`	HTML, Markdown, or parsed JSON — depends on `--format` and `--parse`
`nimble search`	Structured results array (title, URL, description)
`nimble map`	URL list + metadata
`nimble crawl`	Async job — poll with `nimble crawl status <job_id>`

Agent runs always need
--transform "data.parsing"
. If the agent name suggests a list (serp, search, plp), expect an array. If it suggests a single item (pdp, product, profile), expect a dict.

命令	输出
`nimble agent`	`data.parsing` 中的结构化数据——数组（SERP/列表）或字典（PDP/产品）
`nimble extract`	HTML、Markdown或解析后的JSON——取决于 `--format` 和 `--parse` 参数
`nimble search`	结构化结果数组（标题、URL、描述）
`nimble map`	URL列表+元数据
`nimble crawl`	异步任务——使用 `nimble crawl status <job_id>` 轮询状态

运行agent必须使用
--transform "data.parsing"
。若agent名称暗示列表（serp、search、plp），则返回数组。若暗示单个条目（pdp、product、profile），则返回字典。

Output & Organization

输出与组织

bash

mkdir -p .nimble   # save all outputs here

Naming:

.nimble/<site>-<task>.md

(e.g.

.nimble/amazon-airpods.md

.nimble/yelp-sf-italian.json

)

Working with saved files:

bash

wc -l .nimble/page.md && head -100 .nimble/page.md
grep -n "price\|rating" .nimble/page.md | head -30

End every response with:

Source: [URL] — fetched live via Nimble CLI

bash

mkdir -p .nimble   # 所有输出保存至此

命名规则：

.nimble/<site>-<task>.md

（例如

.nimble/amazon-airpods.md

、

.nimble/yelp-sf-italian.json

）

处理已保存文件：

bash

wc -l .nimble/page.md && head -100 .nimble/page.md
grep -n "price\|rating" .nimble/page.md | head -30

每个响应末尾必须添加：

来源: [URL] — 通过Nimble CLI实时获取

Self-Improvement

自我优化

The skill maintains

~/.claude/skills/nimble-web-expert/learned/examples.json

At task start: read the file, scan
```
good[]
```
for
```
url_pattern
```
matches → use documented
```
command
```
/
```
tier
```
as starting point. Scan
```
bad[]
```
→ avoid documented pitfalls.
After presenting results: ask "Were these results what you needed?" → on positive feedback, append to
```
good[]
```
with
```
url_pattern
```
,
```
task
```
,
```
command
```
,
```
tier
```
,
```
notes
```
. On negative feedback, ask "What went wrong?" and append to
```
bad[]
```
with
```
url_pattern
```
,
```
task
```
,
```
issue
```
,
```
avoid
```
,
```
better
```
.
Keep entries concise — 5–10 per site. Only write on real feedback, never speculatively.

本技能维护

~/.claude/skills/nimble-web-expert/learned/examples.json

文件。

任务开始时：读取文件，扫描
```
good[]
```
中的
```
url_pattern
```
匹配项→使用已记录的
```
command
```
/
```
tier
```
作为起点。扫描
```
bad[]
```
→避免已记录的陷阱。
呈现结果后：询问“这些结果符合您的需求吗？”→若反馈积极，将
```
url_pattern
```
、
```
task
```
、
```
command
```
、
```
tier
```
、
```
notes
```
追加至
```
good[]
```
。若反馈消极，询问“哪里出问题了？”并将
```
url_pattern
```
、
```
task
```
、
```
issue
```
、
```
avoid
```
、
```
better
```
追加至
```
bad[]
```
。
条目保持简洁——每个网站5–10条。仅在收到真实反馈时添加，切勿推测性添加。

Guardrails

防护规则

NEVER answer from training data for live prices, current news, or real-time data. If Nimble is unavailable, say so.
NEVER skip Step 0 silently. Even if certain there's no agent, announce the check before running extract/search/map.
NEVER retry the same render tier. If a tier returns empty or blocked, escalate — do not re-run.
NEVER substitute WebFetch for nimble CLI tasks. WebFetch is a fallback for fetching Nimble docs only.
NEVER load reference files speculatively. Only read a reference when the current task explicitly needs it.
Task agents MUST use
run_in_background=False
. See nimble-agent-builder delegation model for the why.
Hard retry limit. On error (not empty content): retry at most 2 times with different flags. After 2 errors, report and stop.
Hard 429 rule. On rate-limit error: stop immediately. Do not retry or switch tiers.

绝不从训练数据中作答实时价格、最新新闻或实时数据。若Nimble不可用，如实告知。
绝不静默跳过步骤0。即使确定没有agent，也要在运行提取/搜索/映射前告知用户已完成检查。
绝不重试同一渲染层级。若某层级返回空内容或被阻止，升级层级——切勿重新运行。
绝不使用WebFetch替代Nimble CLI任务。WebFetch仅作为获取Nimble文档的备用方案。
绝不推测性加载参考文件。仅当当前任务明确需要时才读取参考文件。
任务agent必须使用
run_in_background=False
。详见nimble-agent-builder委托模型了解原因。
严格重试限制。出现错误（非空内容）时：最多使用不同标志重试2次。2次错误后，报告并停止。
严格429规则。遇到速率限制错误时：立即停止。切勿重试或切换层级。

Reference files

参考文件

Load only when needed:

File	Load when
`references/recipes.md`	Need a proven command for a common site (Amazon, Yelp, LinkedIn…)
`references/nimble-agents/SKILL.md`	Step 0 lookup — full agent table (50+ sites)
`references/nimble-extract/SKILL.md`	Extract flags, render tiers, browser actions, network capture, parser schemas
`references/nimble-search/SKILL.md`	Search flags, all 8 focus modes
`references/nimble-map/SKILL.md`	Map flags, response structure
`references/nimble-crawl/SKILL.md`	Full async crawl workflow
`references/nimble-tasks/SKILL.md`	Poll tasks/batches, fetch results — for async, batch, and crawl operations
`references/nimble-extract/browser-investigation.md`	Tier 6 — CSS selector/XHR discovery with browser-use or Playwright
`references/nimble-extract/parsing-schema.md`	Parser types, selectors, extractors, post-processors
`references/nimble-extract/browser-actions.md`	Full browser action types and parameters
`references/nimble-extract/network-capture.md`	Filter syntax, XHR mode, capture+parse patterns
`references/nimble-search/search-focus-modes.md`	Decision tree, mode details, combination strategies
`references/batch-patterns.md`	Parallel bash patterns for 2–5, 6–20, and 20+ inputs
`references/error-handling.md`	Error codes, known site issues, troubleshooting

仅在需要时加载：

文件	加载时机
`references/recipes.md`	需要针对常见网站（Amazon、Yelp、LinkedIn…）的成熟命令时
`references/nimble-agents/SKILL.md`	步骤0查找——完整agent表格（50+个网站）
`references/nimble-extract/SKILL.md`	提取标志、渲染层级、浏览器操作、网络捕获、解析器模式
`references/nimble-search/SKILL.md`	搜索标志、全部8种聚焦模式
`references/nimble-map/SKILL.md`	映射标志、响应结构
`references/nimble-crawl/SKILL.md`	完整异步爬取工作流
`references/nimble-tasks/SKILL.md`	轮询任务/批量、获取结果——用于异步、批量和爬取操作
`references/nimble-extract/browser-investigation.md`	层级6——使用browser-use或Playwright发现CSS选择器/XHR路径
`references/nimble-extract/parsing-schema.md`	解析器类型、选择器、提取器、后处理器
`references/nimble-extract/browser-actions.md`	完整浏览器操作类型及参数
`references/nimble-extract/network-capture.md`	过滤语法、XHR模式、捕获+解析模式
`references/nimble-search/search-focus-modes.md`	决策树、模式详情、组合策略
`references/batch-patterns.md`	针对2–5个、6–20个及20个以上输入的并行bash模式
`references/error-handling.md`	错误代码、已知网站问题、故障排查