使用方法
- 复制安装命令,在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/gaket/talent-agent --skill find-people开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。
find-people
Inputs
- Company + role (from tracker or user).
profile/prefs.yml→alma_mater, plus clubs/co-op program if the user mentions them.
Steps
- Map the team: from the JD, note team/org name (e.g. "ML Platform", "Data Science – Growth").
- Search public sources (web search + fetch):
- Company About/Team/Leadership pages, engineering blog author bylines.
- News/funding announcements (TechCrunch, BetaKit, company press) naming founders/execs.
- GitHub org members and top contributors to company repos.
- Conference talks, podcasts, university news, accelerator pages (YC, Velocity, Creative Destruction Lab).
- Alumni-founder angle: look for founders/early employees from the student's school. Example: Moda is a startup founded by University of Waterloo grads that hires heavily from Waterloo — a Waterloo student should lead with that shared connection. Queries:
"<school>" founder startup,"<school> alumni" "founded", accelerator cohort lists, school entrepreneurship pages (e.g. Velocity for Waterloo, UTEST/Hatchery for UofT).
3b. LinkedIn (opt-in): only if linkedin_browser: true in prefs and a browser tool is available, follow resources/linkedin-browser.md (user logged in, read-only, ≤10 profile opens/session, never Connect or Message). This is the best way to find alumni founders: school People page filtered by "founder". Otherwise skip and suggest the user search LinkedIn manually and paste results.
- For each person record: name, role, why relevant (manager of team / recruiter / alum founder / engineer on stack), public source URL, shared connection, and public contact route (company contact form, careers email, public email on personal site/GitHub). Do not guess emails.
- Rank: (a) alumni founders/engineers, (b) likely hiring manager, (c) university/campus recruiter, (d) engineers on the team.
- Output a short table (≤5 people) and hand off to
draft-outreach.
Rules
- LinkedIn only via the opt-in, read-only mode in
resources/linkedin-browser.md. No bulk scraping, no Connect/Message clicks, stop on any warning. The user can always paste LinkedIn text instead. - Public, professional info only; no personal addresses, phone numbers, or data brokers.
- Cite a URL for every person; mark anything uncertain as "unverified".

