LinkedIn

**Quick judgment:** A practical, browser-based LinkedIn automation skill that is well-aligned with OpenClaw’s browser tools. It’s best viewed as a structured playbook for safely driving LinkedIn through a real browser, rather than a high-level API wrapper. **What it does well:** - Provides concrete browser action patterns for core tasks: checking messages, viewing profiles, searching people, and sending connection requests/messages. - Supports both Chrome Relay and an isolated browser profile, with persistent sessions after login. - Emphasizes safety: explicit user confirmation before sending messages or connection requests, clear rate-limiting guidance (~30 actions/hour), and instructions for handling logouts, rate limits, and CAPTCHAs. **Key risks and limitations:** - **Platform policy risk:** Any automation on LinkedIn may violate their Terms of Service; accounts can be rate-limited, challenged, or banned if behavior appears automated, even with conservative usage. - **UI fragility:** Workflows rely on LinkedIn’s web interface. Layout or URL changes can break brittle selectors or navigation patterns, reducing reliability over time. - **Security/privacy:** The advanced `li_at` cookie method requires handling a powerful session cookie. If mishandled or logged insecurely, it could expose full account access. - **Manual overhead:** Many operations still require the model to construct low-level browser actions; there’s no high-level API for complex workflows (e.g., multi-step campaigns, analytics, or robust lead management). **Best-fit scenarios:** - Users who want an AI agent to **read and summarize** their LinkedIn messages and notifications, then help draft replies which the user explicitly approves. - **Lightweight, supervised outreach** for recruiting, sales, or networking where every message and connection request is reviewed and confirmed by the user. - Researchers or professionals who need to **view and summarize profiles** or search results, but accept occasional breakage when LinkedIn’s UI changes. For high-volume or fully autonomous LinkedIn automation, this skill is not sufficient and would significantly raise ToS and account risk; it is more appropriate for low-volume, high-supervision use.