用 Claude + Apify 自动抓取并评分 LinkedIn 职位

输入简历与个人偏好 → Claude 通过 Apify 连接器抓取 LinkedIn 职位 → 去重过滤后输出评分表格与优先投递清单

给谁

需要批量筛选 LinkedIn 职位、关注签证赞助与薪资底限的求职者

你会得到

一次运行从约 700 条抓取结果中筛选出 10 个高优先级投递目标，并生成带评分的电子表格

你需要

Apify

步骤

1. 上传简历并总结背景

   将简历文本提供给 Claude，让它用恰好 3 行总结你的技术栈、经验年限与职级。

   运行提示词 · 来源原文。Claude 对话输入

   复制提示词

   STEP 1 — Resume
   Summarize my stack, experience and level in exactly 3 lines.

2. 回答偏好问题并等待

   Claude 会依次询问：目标职位名称、美国偏好地点（远程/混合/现场+城市）、职级、是否需要签证赞助、最低底薪（美元）。回答后等待下一步，不要让它继续。

   运行提示词 · 来源原文。Claude 对话输入

   复制提示词

   STEP 2 — Interview (then STOP and wait for my answers)
   Ask me: dream job title(s), preferred US location (remote/hybrid/on-site + cities),
   seniority level, whether I need visa sponsorship, and minimum base salary in USD.

3. 通过 Apify 抓取并过滤职位

   Claude 使用 Apify 连接器运行 LinkedIn 职位抓取。执行 4–7 组关键词搜索（针对你的职位变体），限定美国、过去一个月；然后只保留过去 14 天内发布的职位，并按规则去重与排除。

   运行提示词 · 来源原文。Claude 对话输入 / Apify 连接器配置

   复制提示词

   STEP 3 — Scrape
   Use the Apify connector with a LinkedIn Jobs scraper.
   - Run 4–7 keyword searches (variations of my title) for United States, past month.
   - Keep only jobs whose posted date is within the last 14 days.
   - Dedupe in this order: (a) LinkedIn job ID, (b) same company + same normalized
     title in different cities (merge locations into one row), (c) recruiter or
     aggregator reposts of the same description.
   - If I need sponsorship, EXCLUDE roles whose description says any of:
     "no sponsorship", "will not / cannot / unable to sponsor", "not eligible for
     sponsorship", "not open to visa sponsorship", "without the need for current or
     future sponsorship", "U.S. citizens only", "must be a U.S. citizen",
     "U.S. person", "may not be able to... support H-1B", any security clearance.
   - Do NOT exclude a role just because an equal-opportunity statement mentions
     "citizenship".
   - Label sponsorship as "Offered" ONLY if the text says so explicitly.
     Otherwise "Not stated". Never assume.

4. 生成评分表格

   Claude 输出 .xlsx 格式表格，包含职位、公司、地点、工作类型、薪资、发布日期、申请人数、赞助状态、所需技术栈、匹配分数（0–100）、申请优先级、技能缺口与申请链接。按优先级与分数排序。

   运行提示词 · 来源原文。Claude 对话输入 / 表格生成指令

   复制提示词

   STEP 4 — Spreadsheet (.xlsx)
   Columns: Title, Company, Location, Work Type, Salary (USD), Date Posted,
   Applicants, Sponsorship, Required Stack, Match Score (0–100),
   Apply Priority (High/Medium/Low), Skill Gaps, Apply Link.
   - Salary: base range only. Ignore ranges labeled equity, bonus, stock or sign-on.
   - Applicants: keep LinkedIn's buckets ("<25", "200+").
   - Match Score: title fit (30) + share of the JD stack my resume covers (40)
     + core AI/ML present (10) + years-of-experience fit (20).
   - High = score 72+, posted within 7 days, under 100 applicants,
     and salary band not below my minimum.
   - Sort by priority, then score. Add a "Method" tab explaining every rule.

5. 获取投递建议与简历优化

   Claude 给出前 10 优先投递职位（附一句话原因）、简历中最常缺失的 5 个关键词（带出现次数）、基于简历事实重写的 3 条 bullet，以及从有薪资数据的职位中统计出的同级典型底薪范围。

   运行提示词 · 来源原文。Claude 对话输入

   复制提示词

   STEP 5 — Advice
   Top 10 jobs to apply to first (one line each on why), the 5 keywords most
   often missing from my resume (with counts), 3 rewritten resume bullets using
   only facts from my resume, and typical base salary ranges for my level
   computed from the listings that published one.

STEP 1 — Resume
Summarize my stack, experience and level in exactly 3 lines.

STEP 2 — Interview (then STOP and wait for my answers)
Ask me: dream job title(s), preferred US location (remote/hybrid/on-site + cities),
seniority level, whether I need visa sponsorship, and minimum base salary in USD.

STEP 3 — Scrape
Use the Apify connector with a LinkedIn Jobs scraper.
- Run 4–7 keyword searches (variations of my title) for United States, past month.
- Keep only jobs whose posted date is within the last 14 days.
- Dedupe in this order: (a) LinkedIn job ID, (b) same company + same normalized
  title in different cities (merge locations into one row), (c) recruiter or
  aggregator reposts of the same description.
- If I need sponsorship, EXCLUDE roles whose description says any of:
  "no sponsorship", "will not / cannot / unable to sponsor", "not eligible for
  sponsorship", "not open to visa sponsorship", "without the need for current or
  future sponsorship", "U.S. citizens only", "must be a U.S. citizen",
  "U.S. person", "may not be able to... support H-1B", any security clearance.
- Do NOT exclude a role just because an equal-opportunity statement mentions
  "citizenship".
- Label sponsorship as "Offered" ONLY if the text says so explicitly.
  Otherwise "Not stated". Never assume.

STEP 4 — Spreadsheet (.xlsx)
Columns: Title, Company, Location, Work Type, Salary (USD), Date Posted,
Applicants, Sponsorship, Required Stack, Match Score (0–100),
Apply Priority (High/Medium/Low), Skill Gaps, Apply Link.
- Salary: base range only. Ignore ranges labeled equity, bonus, stock or sign-on.
- Applicants: keep LinkedIn's buckets ("<25", "200+").
- Match Score: title fit (30) + share of the JD stack my resume covers (40)
  + core AI/ML present (10) + years-of-experience fit (20).
- High = score 72+, posted within 7 days, under 100 applicants,
  and salary band not below my minimum.
- Sort by priority, then score. Add a "Method" tab explaining every rule.

STEP 5 — Advice
Top 10 jobs to apply to first (one line each on why), the 5 keywords most
often missing from my resume (with counts), 3 rewritten resume bullets using
only facts from my resume, and typical base salary ranges for my level
computed from the listings that published one.

RULES: Only use real listings returned by the connector. Never invent jobs,
salaries, applicant counts or links. If data is missing, write "Not listed".