job-search for coding agents

AI UNDERDOGSDAILY PICK

AI UNDERDOGS

让 AI 帮你找工作

Give your agent a job hunt

agent-data/job-search

ChatGPT 能帮你改简历

但它不会真的去 LinkedIn 帮你投

这个项目

让 coding agent 真的能替你找工作

ChatGPT can rewrite your resume, but it won't

actually go browse LinkedIn for you. This project

gives your coding agent a real job hunt

skill

Agent 的求职插件

A job-hunt plugin for agents

它不是一个独立 app

而是一套插件

装到你的 coding agent 里

agent 就能自动搜索、筛选

总结职位信息

作者同时维护了四个平台的适配器——Claude Code

Codex、Cursor

Factory——不是只做个 demo 就跑路那种

It's not a standalone app — it's a

plugin you install into your coding agent. Once

set up, the agent can search, filter, and

summarize job listings on its own. The author

ships adapters for four platforms in parallel: Claude

Code, Codex, Cursor, and Factory

★ SIGNAL 1

技能拆成模块

Skills, not one giant prompt

v0.8 把整个架构重构成可组合的技能模块——搜索是一个 skill

筛选是一个，总结又是一个

大多数同类项目就是塞一个巨型 prompt 让模型一口气跑完

这里 agent 自己判断该调哪个技能

这才是 agent 该有的工作方式

v0.8 rewired the whole thing into composable skills

— search, filter, summarize as separate modules. Most

projects dump one giant prompt and hope for

the best. Here the agent decides which skill

to invoke, which is honestly how agents should

work

★ SIGNAL 2

497 次提交

497 commits and counting

GitHub 上 25 颗 star 的项目

497 次提交

这个数字摆出来你就知道了——不是周末写个 demo push 上去就不管了

四个平台的适配器要分别调试

每个 skill 模块要单独迭代

这背后是实打实的工程量

25 stars on GitHub, 497 commits. That ratio

tells you something — this isn't a weekend

throwaway. Maintaining four platform adapters and iterating each

skill module separately takes real, sustained effort

有意思的是

这个项目的 README 里没有「愿景」

没有「改变求职方式」之类的宏大叙事

就是老老实实写怎么装、怎么配

怎么跑

star 虽然不多

但提交频率说明作者自己也在用

不是写完就扔

The README has no manifesto, no vision statement

— just setup instructions and usage. Low star

count, high commit frequency. Looks like the author

is building this for themselves first, not farming

engagement

AI UNDERDOGS

agent-data/job-search

关注 · 每天发现更多 AI 神作

github.com/agent-data/job-search