Job Hunter
End-to-end job search assistant — from finding opportunities to landing interviews.
Quick Start
1. Set up candidate profile
Create a profile JSON for the user. Use the template at {baseDir}/references/profile-template.json as a starting point. Ask the user about:
- - Target roles and seniority level
- Key skills and tools
- Location preferences (cities + remote)
- Salary expectations
- Dealbreakers and excluded companies
- Preferred industries/domains
Save as profile.json in the workspace.
2. Search for jobs
Use the web_search tool with multiple queries to cast a wide net:
CODEBLOCK0
Expand keywords — don't just search one title. See {baseDir}/references/search-strategies.md for keyword expansion patterns.
Alternative: run the search script if Brave API is available:
CODEBLOCK1
3. Evaluate fit
For each job found, run fit analysis:
CODEBLOCK2
Or evaluate manually using this framework:
- - Skill match (40%): Does user have 60%+ of required skills?
- Seniority match (25%): Right level — not over/under qualified?
- Location match (15%): Compatible location or remote?
- Domain match (10%): Preferred industry/domain?
- Red flags (10%): Excluded companies? Dealbreakers?
Score: 🟢 75+ great | 🟡 55-74 good | 🟠 40-54 stretch | 🔴 <40 skip
4. Present results
For each job, present:
- - Role & Company with direct link
- Fit score with color indicator
- Why it's a match (top 3 skill matches)
- Gaps to address (missing skills to highlight as "eager to learn")
- Salary estimate if available
- Recommendation: Apply / Maybe / Skip
Application Support
Cover letters
Read
{baseDir}/references/cover-letter-guide.md for structure and tone guidelines. Generate tailored cover letters that:
- - Reference specific company details (not generic)
- Map user's experience to top 2-3 job requirements
- Include quantified achievements
- Stay under 350 words
Interview prep
Read
{baseDir}/references/interview-prep.md for complete preparation framework. Help with:
- - Company research summaries
- STAR stories for key requirements
- Tailored "tell me about yourself" script
- Salary negotiation talking points
- Questions to ask the interviewer
Salary research
bash {baseDir}/scripts/salary_research.sh "Job Title" "Location"
Cross-reference 3+ sources. In the Netherlands: factor in 8% holiday allowance, possible 13th month, pension.
Daily Brief Format
When running as a scheduled job search brief:
- 1. New opportunities — jobs found in last 24h with fit scores and direct links
- Application status — updates on pending applications
- Action items — what to apply to today, follow-ups due
- Market intel — industry trends, salary movements, hiring patterns
Tracking
Maintain a job tracker with:
- - Company, role, date found, source URL
- Fit score and recommendation
- Status:
new → applied → screening → interview → offer/rejected/ INLINECODE12 - Applied/skipped with reason
- Contact info and follow-up dates
Tips for Agents
- - Never apply on behalf of the user — present opportunities, let them decide
- Don't overwhelm — 3-5 quality matches beat 20 mediocre ones
- Track excluded companies — never suggest the same company twice after rejection
- Be honest about fit — stretches are okay to flag, but don't oversell poor matches
- Respect dealbreakers — if user said no customer service, don't suggest it even if "it's a great company"
- Update the profile — as you learn user preferences, refine the profile
- Celebrate wins — applied to a job? Got an interview? Acknowledge it
求职猎手
端到端求职助手——从发现机会到获得面试。
快速开始
1. 设置候选人档案
为用户创建一份档案JSON文件。以{baseDir}/references/profile-template.json模板为起点。向用户询问以下信息:
- - 目标职位与资历级别
- 关键技能与工具
- 地点偏好(城市+远程)
- 薪资期望
- 否决条件与排除公司
- 偏好的行业/领域
将文件保存为工作区中的profile.json。
2. 搜索职位
使用web_search工具配合多个查询词,广泛撒网:
site:linkedin.com/jobs [职位] [城市]
site:indeed.com [职位] [城市]
site:glassdoor.com/job [职位] [城市]
[职位] [城市] 招聘 2025 2026
扩展关键词——不要只搜索一个职位名称。关键词扩展模式请参见{baseDir}/references/search-strategies.md。
备选方案:如果Brave API可用,运行搜索脚本:
bash
{baseDir}/scripts/search_jobs.sh 客户体验经理 --location 阿姆斯特丹 --days 7
3. 评估匹配度
对每个找到的职位进行匹配度分析:
bash
python3 {baseDir}/scripts/analyze_fit.py --profile profile.json --jobs jobs.json --threshold 50
或使用以下框架手动评估:
- - 技能匹配(40%):用户是否具备60%以上的所需技能?
- 资历匹配(25%):级别合适——不过高或过低?
- 地点匹配(15%):地点兼容或可远程?
- 领域匹配(10%):偏好的行业/领域?
- 警示信号(10%):排除的公司?否决条件?
评分:🟢 75分以上优秀 | 🟡 55-74分良好 | 🟠 40-54分勉强 | 🔴 40分以下跳过
4. 呈现结果
对每个职位,呈现以下信息:
- - 职位与公司(附直接链接)
- 匹配度评分(带颜色标识)
- 匹配原因(前3项技能匹配)
- 需弥补的差距(缺失技能,标注为乐于学习)
- 薪资估算(如有)
- 建议:申请 / 考虑 / 跳过
申请支持
求职信
阅读{baseDir}/references/cover-letter-guide.md了解结构与语气指南。生成量身定制的求职信,需满足:
- - 引用公司具体细节(非通用模板)
- 将用户经验映射到前2-3项职位要求
- 包含量化成就
- 控制在350字以内
面试准备
阅读{baseDir}/references/interview-prep.md获取完整准备框架。提供以下帮助:
- - 公司研究摘要
- 针对关键要求的STAR故事
- 量身定制的自我介绍脚本
- 薪资谈判要点
- 向面试官提问的问题
薪资调研
bash
bash {baseDir}/scripts/salary_research.sh 职位名称 地点
交叉参考3个以上来源。在荷兰:需考虑8%的假期津贴、可能的第13个月工资、养老金。
每日简报格式
作为定时执行的求职简报运行时:
- 1. 新机会——过去24小时内找到的职位,附匹配度评分和直接链接
- 申请状态——待处理申请的更新
- 待办事项——今天需申请的内容、需跟进的后续事项
- 市场情报——行业趋势、薪资变动、招聘模式
追踪管理
维护一份职位追踪表,包含:
- - 公司、职位、发现日期、来源URL
- 匹配度评分与建议
- 状态:新发现 → 已申请 → 筛选 → 面试 → 录用/被拒/无回应
- 已申请/已跳过及原因
- 联系信息与跟进日期
给助手的提示
- - 切勿代用户申请——呈现机会,让用户决定
- 不要信息过载——3-5个优质匹配胜过20个平庸选项
- 记录排除公司——被拒后不再推荐同一家公司
- 诚实评估匹配度——可以标注勉强匹配,但不要过度推销低分选项
- 尊重否决条件——如果用户说不做客服,即使公司很好也不要推荐
- 更新档案——随着了解用户偏好,不断完善档案
- 庆祝进展——申请了职位?获得了面试?及时给予肯定