Token导航 LogoToken导航

Agent Skill · 研究检索

sem(扫描电镜)

sem 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

sem 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 3,099。

总安装 48,888周安装 2,037GitHub Stars 4下载 16,296核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:sem(扫描电镜)
来源仓库:https://github.com/ivangdavila/sem
安装命令:
openclaw skills install sem
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install sem

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

扫描电镜用于查找、检索和筛选相关信息,辅助快速定位候选结果。适合在 OpenClaw 中根据关键词或任务场景使用。搜索引擎营销策略、关键词竞价、质量得分优化和付费搜索活动管理。安装命令:openclaw skills install sem。建议确认权限范围和维护状态,避免触发联网或文件读写。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install sem;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

sem 怎么安装?

在终端执行 openclaw skills install sem。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

sem 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

sem 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

sem 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

sem 用的人多吗?

公开数据是累计安装 48,888 次,近一周安装 2,037 次,下载 16,296 次,GitHub 4 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

sem 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
sem

扫描电镜 · 本页

48,8884需要联网MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
SEM
description
Search engine marketing strategy, keyword bidding, quality score optimization, and paid search campaign management
metadata
category
marketing
skills
["sem", "ppc", "google-ads", "paid-search", "adwords"]

Keyword Strategy

  • Exact match isn't exact anymore — Google includes "close variants" (plurals, misspellings, implied intent). Review search terms report weekly
  • Broad match needs smart bidding — without automated bidding, broad match burns budget on irrelevant queries
  • Long-tail keywords have lower CPC but need volume — 100 low-volume keywords won't spend, consolidate into phrase/broad with negatives
  • Competitor brand bidding is legal but expensive — CPCs are 2-5x higher and Quality Scores stay low. Budget accordingly
  • Single Keyword Ad Groups (SKAGs) are outdated — Google's algorithm works better with theme-based ad groups of 5-15 related keywords

Quality Score Impact

  • Quality Score affects CPC exponentially — QS 10 pays ~50% less than QS 5 for same position
  • Three components: expected CTR, ad relevance, landing page experience. Check diagnostics to know which to fix
  • Historical CTR matters most — new keywords inherit account-level history. Poor performers drag down new campaigns
  • Landing page speed is a ranking factor — compress images, remove unnecessary scripts before scaling spend
  • Ad relevance requires keyword in headline — dynamic keyword insertion ({KeyWord:Default}) is cheap but effective

Bidding Mistakes

  • Manual CPC leaves money on table at scale — switch to automated bidding after 30+ conversions/month
  • Target CPA too aggressive kills volume — start 20% above actual CPA, tighten gradually
  • Maximize Conversions without cap spends entire budget regardless of efficiency — always set a target CPA or ROAS
  • Enhanced CPC (ECPC) is neither manual nor smart — commit to one strategy, ECPC is the worst of both
  • Bid adjustments stack multiplicatively — +50% device × +30% location = +95%, not +80%

Campaign Structure

  • Search and Display in same campaign is always wrong — different intents, different metrics, split them
  • Brand campaigns should be separate — protect branded terms, track true branded vs non-branded performance
  • Geographic targeting at campaign level, not ad group — easier budget control by region
  • Audience layering (observation mode) first — collect data before switching to targeting mode that restricts reach
  • Shared budgets hide which campaigns would spend more — use individual budgets to understand true demand

Ad Copy Rules

  • Include primary keyword in Headline 1 — ad relevance and CTR both improve
  • Responsive Search Ads need 8-10 headline variations — algorithm can't optimize with only 3-4 options
  • Pin critical messaging to position 1 — brand name, key offer, or compliance text that must always show
  • Call extensions increase CTR 5-10% — add phone number even if calls aren't primary goal
  • Test exactly one element at a time — headline vs headline, not headline + description + CTA simultaneously

Search Terms Hygiene

  • Review search terms weekly, not monthly — irrelevant spend compounds fast
  • Add negatives at campaign level for broad exclusions, ad group level for specific refinement
  • Negative keyword match types matter — "free" as phrase negative blocks "free trial" but not "trial free download"
  • Create a shared negative keyword list — apply learnings across campaigns automatically
  • Search term "other" category is hidden queries — if it's significant %, request detailed report from rep or use scripts

Conversion Tracking Errors

  • Last-click attribution hides assist value — at minimum, compare with data-driven attribution
  • Counting "All conversions" inflates ROAS — filter to primary conversion action for true performance
  • Phone call conversions need duration minimums — 60 seconds minimum filters spam and misdials
  • Micro-conversions (email signups) have different values than purchases — set values proportionally
  • Cross-device conversions are modeled, not tracked — factor uncertainty into ROAS calculations

Scaling Paid Search

  • Search volume has a ceiling — unlike social, you can't pay to reach more people than are searching
  • Impression Share shows headroom — if IS is 90%+, growth requires new keywords, not more budget
  • Diminishing returns are real — CPAs increase 20-40% when scaling past optimal volume. Know your efficiency/volume tradeoff
  • Bing Ads imports Google campaigns directly — 10-15% extra volume at lower CPCs, minimal extra work
  • International expansion multiplies complexity — separate accounts per market, don't just add countries to existing campaigns

Platform Automation

  • Auto-applied recommendations are opt-out, not opt-in — disable recommendations that conflict with strategy
  • Smart campaigns hide too much data — use regular campaigns for control and learning
  • Performance Max cannibalizes branded search — exclude brand terms or track incrementality carefully
  • Scripts automate repetitive tasks — bid adjustments, budget pacing, negative keyword mining run automatically
  • Rules-based automation is fragile — "if CTR < 2%, pause" kills potentially good ads too early. Use with caution

平台分布

OpenClaw

94.03%
按下载量换算15,323

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

未展示

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills