recruiting-copilot skill
Skills for HR and recruiters: role requirements, sourcing on BOSS Zhipin and Liepin, resume review, interviews and a candidate ledger
- 109 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent signs in to your employer account and can write to people in your name, so a mistaken send reaches real candidates
- The ledger and archive hold candidates' personal data: names, resumes and scores sit on your disk
- Automated actions on BOSS Zhipin and Liepin may break the services' rules and lead to an account restriction
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/recruiting-copilotDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent signs in to your employer account and can write to people in your name, so a mistaken send reaches real candidates
- The ledger and archive hold candidates' personal data: names, resumes and scores sit on your disk
- Automated actions on BOSS Zhipin and Liepin may break the services' rules and lead to an account restriction
This is third-party code. Review the repository files before installing.
What it does
recruiting-copilot is a set of skills and plugin commands for recruiting in Claude Code, Codex and other agents that read project files. The `recruit-init` skill checks dependencies and creates a separate working folder with templates. Then `recruit-grill` asks one question at a time to find out whom you are hiring, and turns the answers into a public job description, internal sourcing notes, hard rules and the first search queries. The daily skill `recruit-daily` reads the latest requirements, checks unread replies on BOSS Zhipin and Liepin, searches both services for candidates, screens them against your criteria and records evidence and weak points. The agent contacts candidates only with your permission for the current pass and reports every send. The ledger `02-sourcing/dedup-ledger.csv` removes duplicates by name and role, and a daily report closes the pass. The other skills cover neighbouring tasks: `resume-review` scores resumes that arrive from headhunters, referrals or directly; `market-talent-mapping` surveys the talent market for a role; `interview-schedule` books interviews with a video call and interviewers through the optional lark-cli. A separate plugin for DeepSeek Harness adds a progress board.
Who it is for. HR specialists and headhunters who source people on the Chinese platforms BOSS Zhipin and Liepin and want to keep criteria and the ledger on their own disk rather than in someone else's service.
Good fit when
- A new role opens and the requirements still read "we need a strong person"
- Every morning you have to go through replies on BOSS Zhipin and Liepin and find new candidates against the same criteria
- A headhunter sent a pile of resumes and they have to be compared on one scale
- Before hiring you want to know how many suitable people the market has at all
- A candidate reached the interview stage and you have to agree on time with interviewers and create a video call
Not a fit when
- Hiring happens on other platforms: only BOSS Zhipin and Liepin are built in
- You have no employer account on those services: without a login the search will not start
- The agent should message people on its own without your approval: contact waits for your confirmation by default
- You need documentation in Russian or English: the detailed instructions in the repository are in Chinese
Example request
Run today's recruiting: check new replies on BOSS Zhipin and Liepin, find suitable candidates for the backend developer role and prepare the report, do not message anyoneLimitations
The detailed documentation and skill texts are in Chinese; the project description is partly bilingual. Sourcing needs Node.js 22 or newer, Chrome or Edge and your own login to BOSS Zhipin or Liepin, done by hand in an opened window with `boss login` and `liepin login`. The installer pulls the third-party packages `@viyzhu/boss-cli-fork` and `@viyzhu/liepin-cli`; the source can be swapped with `BOSS_CLI_SOURCE` and `LIEPIN_CLI_SOURCE`. Automating actions in someone else's service may break its rules, and the account stays your responsibility. Calendar and video calls work only after lark-cli is set up; without it the agent returns a list for manual creation.
How to disable. Disable or remove the plugin through `/plugin` in Claude Code or delete the skill folders from the project. The workspace with the ledger and candidate files sits elsewhere; delete it by hand. Log out of the accounts in the browser to close the service sessions.
FAQ
How do I install recruiting-copilot in Claude Code?
Clone the repository with `git clone --branch v0.8.0 --depth 1 https://github.com/Viy1204/recruiting-copilot.git`, add it as a plugin source with `claude plugin marketplace add` and an absolute path, then run `claude plugin install recruiting-copilot@recruiting-copilot`. After restarting the session, start with `/recruiting-copilot:recruit-init`.
Does the agent message candidates by itself?
No, not until you allow it. By default the greeting waits for your confirmation. If you say "message the suitable ones right away", the permission covers only the current pass and does not carry over to the next day. If a send cannot be verified, the person is not recorded as contacted.
Where is the candidate data kept?
In a separate workspace that `recruit-init` creates: role criteria, the ledger `02-sourcing/dedup-ledger.csv`, interview files and reports. The repository itself holds only code and templates, so you can update it without touching the data.
Do I need an employer account?
Yes, sourcing and replies need a login to BOSS Zhipin or Liepin. You do it by hand: `boss login` or `liepin login` opens a browser window for a code scan. Without a login only resume review and role scoping work.
Does the set work in Codex?
Yes, the authors name Codex, WorkBuddy, Qoder, MiniMax Code and ZCode. The repository holds copies of the skills for the `.agents/skills`, `.claude/skills` and `.qoder/skills` folders. Commands like `/recruiting-copilot:recruit-daily` exist only in the Claude Code plugin; in other agents say it in words: "run today's recruiting".
Discussion
No questions or reviews yet.
Related
SkillsA Claude Code skill that collects BOSS Zhipin vacancies, matches them against a resume and can apply to the best ones.
An agent on your Claude or Codex subscription finds jobs, tailors resumes and applies, with every status shown on a web dashboard.
Skill and CLI for job applications: fills Greenhouse, Ashby and Lever forms and logs only confirmed submissions
An MCP server and CLI for the BOSS job platform: search vacancies, filter by conditions and shortlist with JSON output
A skill and SQLite engine that watch thousands of sources on a schedule, detect changes by hash and keep an audit ledger.
Eight open source AI employees for business roles: 60 scheduled routines, work in your own browser and a morning brief, on Claude Code and other agents
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| recruiting-copilot (this card) | no | 109 | ||
| Huntflow MCP server | no | API key | 1 | connects the agent to the Russian recruiting system Huntflow rather than Chinese platforms |
| HeadHunter MCP Server | no | OAuth | 36 | works with the HeadHunter site through MCP, without a ledger or a daily loop |
README in short
The Chinese description opens with the idea: first the real requirements for each role, then sourcing on two platforms, screening and permitted contact, finally a ledger and a report. Next come the pinned-version install, service logins, the Claude Code plugin, a DeepSeek Harness variant, optional Feishu integration and a table of every skill with its main outputs.
README badge
Original SKILL.md
---
name: recruit-daily
description: >
每日招聘流水线：把"查 boss+猎聘未读 → 双通道主动寻源 → 硬规则初筛 → 打招呼 → 补台账 → 出日报"
串成一条流水线。当用户说"处理今天的招聘"、"有没有未处理人选"、"帮我找找 X 岗的人"、
"给合适的人打招呼"、"出今天的招聘日报"，或在招聘工作区做任何寻源/初筛/建档动作时使用。
---
# recruit-daily —— 每日招聘流水线
每日主力动作：把 Boss 直聘 + 猎聘两个渠道的「查未读 → 主动寻源 → 初筛 → 打招呼 → 建档 → 出日报」串成一条流水线。
网页开工（默认）：确认用户招聘工作区后，按 [`../recruit-board/references/hosts.md`](../recruit-board/references/hosts.md) 统一接入；脚本可用时为 `daily` 先 start、保存 run id 并进入真实步骤，在耗时寻源前确认服务并打开或复用网页，同时交付本轮链接。步骤、降级和记录命令遵循 [`共享进度协议`](../recruit-board/references/progress.md)，按下方完成判据结束各步，跳步写理由，等待授权记 `wait`。
在 unread / screen / greet 等实际步骤中每处理完一人就及时写 `result`，初筛未通过、失败或跳过也写理由，未核实发送不记成功；待确认名单先记 `decision` 再按原安全边界 `wait`，用户拍板后按共享协议更新原收据消解待确认，不等日报结束才补录。
**这是编排文档**：CLI 命令细节见 [`references/channels.md`](references/channels.md)，
台账与日报格式见 [`references/ledger-and-report.md`](references/ledger-and-report.md)，
日报卡片推送见 [`references/daily-card.md`](references/daily-card.md)。
本文档只管「顺序、判断、跨渠道协同、安全」。
## 铁律：先读事实源，别凭记忆
开工**先读**（都在工作区根下）：
- `CONTEXT.md` —— 术语、**初筛硬规则**、**在招岗位与优先级**（唯一事实源，随时可能更新）。
- `AGENTS.md` —— 分工与「不可逆动作等确认」红线。
- `01-jd/_internal/<role>.md` —— 各岗命脉技能 / 排除信号 / **关键词迭代表**（寻源前必读，越搜越准）。
**本文档不写任何硬规则数字**——年龄线、学历线、地点要求全部以 CONTEXT 为准，避免两处打架。
如果 CONTEXT 硬规则区还是空的（没跑过岗位梳理），停下来先走 `skills/recruit-grill/SKILL.md`。
## 主流程（按需跳步）
1. **查未处理**：
- `boss list --unread`（角标 > 0 才有）
- `liepin chatlist --json`，过滤 `unread_count` 非 "0"。
- **有未读**：逐个按 CONTEXT 硬规则判 → 该回复的回复、该要简历的要、
该排除的**先记台账等用户确认**（绝不自动点"不合适"）。
- **没未读**：转主动寻源。
2. **定寻源岗位与优先级**：读 CONTEXT「在招岗位与优先级」，按优先级分配今天的额度。
3. **双通道 × 双入口寻源**（每个目标岗都跑），命令与坑见 channels.md：
- **boss**：人才库 `boss search <词> --job <岗>` + 推荐 `boss recommend <岗>`。
（`boss positions` 当前读不出职位，别拿它拿岗位槽位——岗位名直接用 CONTEXT 里的。）