diff --git a/agent_reach/skill/references/career.md b/agent_reach/skill/references/career.md
index f1b46fe7..c09124ae 100644
--- a/agent_reach/skill/references/career.md
+++ b/agent_reach/skill/references/career.md
@@ -1,29 +1,95 @@
-# 职场招聘
+# 职场招聘 / Career
 
-LinkedIn。
+LinkedIn — via the `linkedin-scraper-mcp` server (browser automation), invoked
+through mcporter. It is registered as the mcporter server `linkedin`, so every
+call is `mcporter call 'linkedin.<tool>(...)'`.
 
-## LinkedIn
+## Setup
 
 ```bash
-# 获取个人资料
-mcporter call 'linkedin-scraper.get_person_profile(linkedin_url: "https://linkedin.com/in/username")'
+# One-time interactive login (opens a browser; session saved to ~/.linkedin-mcp/profile)
+linkedin-scraper-mcp --login --no-headless
+linkedin-scraper-mcp --status   # verify the session is valid
 
-# 搜索人才
-mcporter call 'linkedin-scraper.search_people(keyword: "AI engineer", limit: 10)'
+# Register with mcporter (stdio transport; no long-running server needed)
+mcporter config add linkedin --command linkedin-scraper-mcp --arg --transport --arg stdio
+```
+
+> Requires a valid login. Without it, only the Jina fallback (public pages) works.
+> Run `mcporter list linkedin` to see the live tool signatures.
+
+Conventions: people are identified by **username** (`"williamhgates"`, not a URL),
+companies by their **URL slug** (`"anthropicresearch"`, not the display name), jobs
+by **`job_id`**. Search params are `keywords` (plural).
+
+## People
+
+```bash
+# A person's profile. sections (comma-separated, optional): experience, education,
+# interests, honors, languages, certifications, skills, projects, contact_info, posts
+mcporter call 'linkedin.get_person_profile(linkedin_username: "williamhgates", sections: "experience,education")'
 
-# 获取公司资料
-mcporter call 'linkedin-scraper.get_company_profile(linkedin_url: "https://linkedin.com/company/xxx")'
+# The authenticated user's own profile
+mcporter call 'linkedin.get_my_profile(sections: "experience,skills")'
 
-# 搜索职位
-mcporter call 'linkedin-scraper.search_jobs(keyword: "software engineer", limit: 10)'
+# Search people. network: ["F"]=1st, ["S"]=2nd, ["O"]=3rd+. current_company takes a
+# numeric company URN id (from get_company_profile), not a plain company name.
+mcporter call 'linkedin.search_people(keywords: "AI engineer", location: "Remote", network: ["F"])'
+
+# Profiles from the "people you may know" sidebar of a profile page
+mcporter call 'linkedin.get_sidebar_profiles(linkedin_username: "williamhgates")'
 ```
 
-> **需要登录**: LinkedIn scraper 需要有效的登录态。
+## Companies
+
+```bash
+# Company "about" page. sections (optional): posts, jobs
+mcporter call 'linkedin.get_company_profile(company_name: "anthropic", sections: "posts")'
+
+# Recent posts from a company feed
+mcporter call 'linkedin.get_company_posts(company_name: "docker")'
+
+# Search companies
+mcporter call 'linkedin.search_companies(keywords: "fintech")'
 
-### Fallback 方案
+# Employees + demographics (location / education / function breakdown).
+# company_name MUST be the exact URL slug, e.g. "anthropicresearch" not "anthropic".
+mcporter call 'linkedin.get_company_employees(company_name: "anthropicresearch", keywords: "engineer")'
+```
 
-如果 MCP 不可用，可以用 Jina Reader：
+## Jobs
 
 ```bash
-curl -s "https://r.jina.ai/https://linkedin.com/in/username"
+# Search jobs -> returns job_ids. Optional filters:
+#   max_pages 1-10 (default 3) | date_posted: past_hour|past_24_hours|past_week|past_month
+#   job_type: full_time,part_time,contract,temporary,volunteer,internship,other
+#   experience_level: internship,entry,associate,mid_senior,director,executive
+#   work_type: on_site,remote,hybrid | easy_apply: true|false | sort_by: date|relevance
+mcporter call 'linkedin.search_jobs(keywords: "software engineer", location: "Remote", max_pages: 1, work_type: "remote")'
+
+# Full detail for one posting (job_id from search_jobs)
+mcporter call 'linkedin.get_job_details(job_id: "4252026496")'
+```
+
+## Feed & messaging (read)
+
+```bash
+mcporter call 'linkedin.get_feed(num_posts: 10)'               # authenticated home feed
+mcporter call 'linkedin.get_inbox(limit: 20)'                  # messaging inbox
+mcporter call 'linkedin.get_conversation(linkedin_username: "williamhgates")'
+mcporter call 'linkedin.search_conversations(keywords: "interview", limit: 10)'
 ```
+
+> The MCP also exposes write tools (`connect_with_person`, `send_message`) plus
+> `close_session`. Writes are out of scope for this read/search skill — only use
+> them on explicit user instruction.
+
+## Fallback (no login, public pages only)
+
+```bash
+curl -s "https://r.jina.ai/https://linkedin.com/in/williamhgates"
+```
+
+---
+Note: the pip package now prints a rename notice — it is `mcp-server-linkedin`
+(`uvx mcp-server-linkedin@latest`); the `linkedin-scraper-mcp` entrypoint still works.