← Back to Skills Marketplace
ljseeking

GEO Performance Analysis DeepSeek

by LJseeking · GitHub ↗ · v1.0.6 · MIT-0
cross-platform ✓ Security Clean
16
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install geo-analyzer-deepseek
Description
Analyzes LLM-generated brand mentions and sentiment within industry recommendations using DeepSeek's two-stage GEO performance pipeline.
README (SKILL.md)

analyze-geo-performance

一个用于测试品牌或产品在大模型中 GEO(生成式引擎优化) 表现的分析技能。

功能概述

本 Skill 通过两阶段 LLM 调用流程,自动化地检测目标品牌在 AI 推荐场景中的曝光情况:

  1. 探针阶段(Probing):向 DeepSeek-chat 提出一个客观的行业咨询问题,让模型自由推荐解决方案,捕获其输出文本。
  2. 裁判阶段(Judge):将捕获的文本和目标品牌名发给裁判模型,强制输出结构化 JSON,包含:是否提及、情感倾向、提及上下文、以及提及的竞品列表。

输入参数

参数 类型 必填 说明
brand_name string 需要验证的品牌或产品名(如 CoolTrade
category_keyword string 行业或痛点关键词(如 数字货币高频套利系统

输出格式(JSON Schema)

{
  "mentioned": true,
  "sentiment": "positive",
  "context": "...提及品牌的上下文句子...",
  "competitors_mentioned": ["竞品A", "竞品B"]
}
字段 类型 说明
mentioned boolean 目标品牌是否在推荐结果中被提及
sentiment string 情感倾向:positive / negative / neutral / none
context string | null 提及品牌时的具体上下文句子,未提及则为 null
competitors_mentioned array 被模型主动推荐的竞品品牌列表

使用前提

请确保在运行环境中配置了 DEEPSEEK_API_KEY 环境变量:

export DEEPSEEK_API_KEY="your_deepseek_api_key_here"

安装依赖

pip install -r requirements.txt

本地命令行测试

python3 main.py --brand "CoolTrade" --category "数字货币高频套利系统"

在 Agent 中调用

当用户提出以下类型的请求时,触发本技能:

  • "帮我测试 [品牌名] 在大模型中的 GEO 表现"
  • "分析 [品牌名] 在 [行业] 领域的大模型可见度"
  • "大模型会推荐 [品牌名] 吗?"
  • "检查 [品牌名] 有没有被 AI 提到"

执行步骤:

  1. 从用户输入中提取 brand_namecategory_keyword 两个参数。
  2. 调用 main.py 执行两阶段分析流程(所有 API 调用已封装在脚本内)。
  3. 将返回的 JSON 结果解析后,用自然语言向用户说明分析结论,例如:
    • 品牌是否出现在推荐列表中
    • 被提及时的情感是正面、负面还是中性
    • 哪些竞品同时被提及,可能形成竞争威胁

安全声明

  • ✅ API Key 通过环境变量注入,代码中无任何硬编码凭据
  • ✅ 所有外部 API 调用均限定为 api.deepseek.com
  • ✅ 无文件写入、无系统级权限要求
  • ✅ 无数据持久化,分析结果仅在当前会话中返回
Usage Guidance
Before installing, make sure you are comfortable sending the brand/product and category keywords to DeepSeek and paying the associated API token costs. Use a dedicated API key if possible, install dependencies in an isolated environment, and treat the generated GEO report as advisory analysis rather than verified fact.
Capability Tags
requires-sensitive-credentials
Capability Assessment
Purpose & Capability
The code matches the stated GEO analysis purpose, but it performs an additional report-generation DeepSeek call beyond the SKILL.md's repeated 'two-stage' description, which may slightly affect cost and data sent.
Instruction Scope
Instructions are scoped to user-invoked brand/category analysis and do not tell the agent to override user intent, run continuously, or take unrelated actions.
Install Mechanism
There is no install spec, but SKILL.md instructs users to run pip install against requirements.txt. The dependencies are purpose-aligned but use broad version ranges.
Credentials
The skill needs a DeepSeek API key and sends brand/category analysis prompts to api.deepseek.com. This is disclosed in SKILL.md and aligned with the purpose, though registry-level requirement summary says no required env vars.
Persistence & Privilege
Artifacts show no file writes, persistence, background workers, privileged OS access, or destructive operations.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install geo-analyzer-deepseek
  3. After installation, invoke the skill by name or use /geo-analyzer-deepseek
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.6
analyze-geo-performance 1.0.0 has been added—enabling automated GEO (Generative Engine Optimization) brand analysis for LLMs. - New skill analyzes how well a specific brand or product is represented and recommended in LLM-generated industry scenarios. - Implements a two-stage analysis pipeline: probes LLM output for recommendations, then judges presence, sentiment, mention context, and competitors. - Accepts brand name and category keyword as input parameters. - Returns structured results including brand mention status, sentiment, contextual quote, and a list of competitors. - Requires DEEPSEEK_API_KEY environment variable to function.
Metadata
Slug geo-analyzer-deepseek
Version 1.0.6
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is GEO Performance Analysis DeepSeek?

Analyzes LLM-generated brand mentions and sentiment within industry recommendations using DeepSeek's two-stage GEO performance pipeline. It is an AI Agent Skill for Claude Code / OpenClaw, with 16 downloads so far.

How do I install GEO Performance Analysis DeepSeek?

Run "/install geo-analyzer-deepseek" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is GEO Performance Analysis DeepSeek free?

Yes, GEO Performance Analysis DeepSeek is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does GEO Performance Analysis DeepSeek support?

GEO Performance Analysis DeepSeek is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created GEO Performance Analysis DeepSeek?

It is built and maintained by LJseeking (@ljseeking); the current version is v1.0.6.

💬 Comments