AI Job Hunter Pro
Intelligent job search assistant with RAG-based semantic matching, automated applications, and data-driven tracking.
Setup (first-time only)
Run the setup script to install dependencies and initialize the vector database:
CODEBLOCK0
Then create your profile:
CODEBLOCK1
Import your resume (PDF or DOCX):
CODEBLOCK2
Core Workflows
Workflow 1: Smart Job Search (RAG Matching)
When user says "find jobs for me" or "match my resume":
- 1. Load user profile from INLINECODE0
- Run RAG matching engine:
python3 {SKILL_DIR}/scripts/rag_engine.py \
--mode search \
--platforms linkedin,boss \
--min-score 0.75 \
--max-results 20
- 3. Present results sorted by match score
- For each job, show: title, company, match score, top matching skills, missing skills
- Ask user which jobs to apply to, or auto-apply if configured
Workflow 2: Auto-Apply Pipeline
When user says "apply to these jobs" or "auto-apply":
- 1. For each selected job:
python3 {SKILL_DIR}/scripts/apply_pipeline.py \
--job-id <id> \
--mode dry-run \
--generate-cover-letter \
--optimize-ats
- 2. In dry-run mode: show generated cover letter and ATS-optimized resume highlights for review
- After user confirms, switch to INLINECODE1
- Log result to tracker database
Workflow 3: Application Tracking
When user says "track my applications" or "job search report":
CODEBLOCK5
Status flow: Discovered → Applied → Screening → Interview → Offer / Rejected
Workflow 4: Feedback Loop
When user says "I like this job" or "not interested":
CODEBLOCK6
This adjusts the RAG query vectors to improve future recommendations.
Rules
- - Always start in dry-run mode. Never submit applications without explicit user confirmation.
- Respect platform rate limits: max 20 applications per day across all platforms.
- Never misrepresent the user's qualifications in cover letters or applications.
- Store all data locally. Never send resume data to external services other than the job platforms themselves.
- When a platform returns an error or blocks access, report it clearly and suggest manual fallback.
- Always show the match score and reasoning before applying.
Configuration
User config lives at ~/job_profile.json. Skill config in OpenClaw:
CODEBLOCK7
AI Job Hunter Pro
基于RAG语义匹配的智能求职助手，支持自动投递和数据驱动追踪。
安装（仅首次）
运行安装脚本以安装依赖并初始化向量数据库：
bash
cd {SKILL_DIR}
pip install -r scripts/requirements.txt
python3 scripts/setup_rag.py --init
然后创建您的个人资料：
bash
cp assets/profiletemplate.json ~/jobprofile.json
编辑 ~/job_profile.json 填写您的信息
导入您的简历（PDF或DOCX格式）：
bash
python3 scripts/rag_engine.py --import-resume ~/path/to/resume.pdf
核心工作流程
工作流程1：智能职位搜索（RAG匹配）
当用户说帮我找工作或匹配我的简历时：
- 1. 从 ~/job_profile.json 加载用户资料
- 运行RAG匹配引擎：
bash
python3 {SKILL
DIR}/scripts/ragengine.py \
--mode search \
--platforms linkedin,boss \
--min-score 0.75 \
--max-results 20
- 3. 按匹配分数排序展示结果
- 对每个职位显示：职位名称、公司、匹配分数、最匹配的技能、缺失技能
- 询问用户要投递哪些职位，或根据配置自动投递
工作流程2：自动投递流程
当用户说投递这些职位或自动投递时：
- 1. 对每个选定的职位：
bash
python3 {SKILL
DIR}/scripts/applypipeline.py \
--job-id
\
--mode dry-run \
--generate-cover-letter \
--optimize-ats
- 2. 在预览模式下：显示生成的求职信和ATS优化后的简历亮点供审核
- 用户确认后，切换到 --mode submit
- 将结果记录到追踪数据库
工作流程3：投递追踪
当用户说追踪我的投递或求职报告时：
bash
python3 {SKILL_DIR}/scripts/tracker.py --report daily
状态流程：发现 → 已投递 → 筛选中 → 面试 → 录用 / 拒绝
工作流程4：反馈循环
当用户说我喜欢这个职位或不感兴趣时：
bash
python3 {SKILLDIR}/scripts/ragengine.py \
--mode feedback \
--job-id \
--signal like|dislike
这将调整RAG查询向量以改进未来的推荐。
规则
- - 始终以预览模式启动。未经用户明确确认，绝不提交投递。
- 遵守平台速率限制：所有平台每天最多投递20个职位。
- 绝不在求职信或投递中歪曲用户的资质。
- 所有数据本地存储。除职位平台本身外，绝不将简历数据发送给外部服务。
- 当平台返回错误或阻止访问时，清晰报告并建议手动后备方案。
- 在投递前始终显示匹配分数和理由。
配置
用户配置位于 ~/job_profile.json。技能配置在OpenClaw中：
json
{
skills: {
ai-job-hunter-pro: {
enabled: true,
profilepath: ~/jobprofile.json,
default_platforms: [linkedin, boss],
maxdailyapplications: 20,
minmatchscore: 0.75,
require_confirmation: true,
dry_run: true
}
}
}
