LinkedIn Automation Skill
Drive a real cloud browser via CLI so AI agents can search, connect, message, and post on LinkedIn end to end.
Docs transparently disclose data flow (real cloud browser operating LinkedIn), token setup and error codes; no malware or covert exfiltration found. But no user-confirmation requirement for high-risk write actions (send messages, remove connections, create posts, send invites), token storage/rollback unexplained, and a mandatory LINKEDAPI_CLIENT attribution prefix (telemetry) is embedded — deducted to 12.
Docs are self-consistent with coherent examples, exit codes and JSON output; failure feedback is well designed. But static review gives no execution evidence, no tests covering the skill's key paths, and availability of the external CLI/cloud API is unverifiable — deducted to 8 (static cap 10).
Scenarios and trigger conditions are clear, boundaries disclosed (Sales Navigator subscription, 30s–minutes latency). But core function depends entirely on LinkedIn and linkedapi.io, unreachable from mainland-China networks, with no Chinese-language support — deducted to 5.
Well-layered docs (install, auth, commands, behavior), MIT license and npm versioning present, maintenance workflows exist. But no changelog, no known-limitations/FAQ section, maintenance ownership only implicit — deducted to 8.
Claims to fetch, search, message etc. with directly usable structured JSON output; but results unverifiable statically, value gated behind a paid third-party service, and comparative benefit evidence is thin — deducted to 5 (static cap 7).
CI workflows exist (syntax check and installer smoke test), primary material auditable; but CI does not cover the skill's key paths, no test suite, no third-party execution evidence, no cross-source corroboration — deducted to 4 (static cap 5).
- The skill exposes external write actions (send messages, remove connections, create posts) with no documented user-confirmation step; an agent could cause irreversible external effects unsupervised.
- All operations run through Linked API's cloud browser with a mandatory LINKEDAPI_CLIENT attribution prefix; usage telemetry is disclosed to a third party and tokens are stored by the local CLI.
- Core function depends on LinkedIn and linkedapi.io, effectively unreachable from mainland-China networks; Chinese users need accessible networking and should weigh LinkedIn ToS risk for automation.
- LinkedIn accounts risk restriction or suspension from automated behavior; all claimed outcomes are unverified — this is a static documentation review.
What does this skill do, and when should you use it?
This is the general-purpose LinkedIn skill from the Linked-API/linkedin-skills repository. It wraps the linkedin CLI (@linkedapi/linkedin-cli); every command calls Linked API, which runs a real cloud browser against LinkedIn. Capabilities span profile and company fetching, people/company search, messaging, connection management, posting and engagement, statistics, and Sales Navigator. Operations are not instant — expect 30 seconds to several minutes each. Using it requires a Linked API account and two tokens.
Exposes a linkedin CLI that can: fetch person profiles with optional experience, education, skills, languages, posts, comments, and reactions; search people and companies with filters like location, industry, company, and size; fetch companies with employees, decision makers, and posts; send messages and read conversations; check, send, accept, ignore, withdraw, and remove connections; fetch posts, create posts (with image/video/document attachments or as a company page), react, and comment; view SSI, performance analytics, API usage, and profile viewers; use Sales Navigator for lookups, filtered searches, and InMail (requires subscription); and run custom workflow definitions. All output is JSON, with exit codes distinguishing API success from infrastructure errors.
- A salesperson has the agent search LinkedIn for prospects by position, company, and location, then fetch full profiles.
- An SDR checks a contact's connection status, sends a connection request with a note, and follows up with messages.
- A marketing team publishes posts with attachments, comments, and reactions as an individual or on behalf of a company page.
- A recruiter or BD person reviews who viewed their profile and analyzes SSI and content performance stats.
- A team with Sales Navigator needs to search for company decision makers by revenue and experience and send InMails.
What are this skill's strengths and limitations?
- Comprehensive coverage: profiles, search, messaging, connections, posting, stats, Sales Navigator, and custom workflows in one skill.
- Structured JSON output plus clearly documented exit codes make programmatic success/error handling easy for agents.
- Thorough documentation with rich command examples, field selection, and workflow recovery via workflowId.
- Powered by an official CLI and cloud browser — no local browser emulation needed.
- Operations are not instant (30 seconds to several minutes) and run sequentially per account, so bulk tasks are slow.
- Depends on the paid Linked API service (exit code 3 signals a required plan); pricing is not documented in the source.
- Requires authorizing your LinkedIn account to a third-party cloud-browser service, with account-restriction and privacy implications.
- Sales Navigator commands additionally require a LinkedIn Sales Navigator subscription.
- No test suite or cross-platform validation evidence is provided in the source.
How do you install this skill?
Option one: run npx @linkedapi/skills — the installer detects your agents and guides setup; or install just this skill with npx @linkedapi/skills add linkedin --yes. Prerequisites: Node.js ≥ 20, @linkedapi/linkedin-cli installed globally, and a Linked API account at app.linkedapi.io with your LinkedIn account connected to obtain the Linked API Token and Identification Token. Manual fallback: copy the linkedin skill folder into your agent's skills directory (e.g. .claude/skills/linkedin/ or ~/.claude/skills/linkedin/). The source does not document per-platform auto-install details.
How do you use this skill?
Authenticate first: linkedin setup --linked-api-token=TOKEN --identification-token=TOKEN. Then run all commands with -- -q and the LINKEDAPI_CLIENT=skill:linkedin prefix. Examples: linkedin person fetch https://www.linkedin.com/in/username --experience -- -q for a profile with work history; linkedin person search --term "product manager" --locations "San Francisco" -- -q to search people; linkedin message send <url> 'your message' -- -q to message. A trigger prompt for your agent could be: "Use the linkedin skill to find product managers in San Francisco and fetch profiles for the top 5." Note operations queue sequentially and take 30 seconds to several minutes each.
How does this skill compare with similar options?
The same repository also ships a linkedin-growth skill focused on a two-phase lead pipeline (import and qualify leads against your ICP, then send connection invites on a schedule across accounts). This linkedin skill is the general-purpose, per-operation tool and does not include that automated outbound cadence.