linkedin_skills

Build a voice profile from real messages the user has written. Produces a style guide Claude follows every time it drafts outreach. Use when setting up the system for the first time, when reply rates drop, or when the user says their messages sound off. Trigger on: "calibrate my voice", "my messages sound like a bot", "build my style guide", or any time the user pastes a set of messages they've personally written. Run once per quarter or whenever your communication style shifts.

Filter a CSV of LinkedIn leads against ICP criteria and persist canonical lead state. Outputs qualified/disqualified files, a contract registry JSON, and weekly metrics. Use when you have a lead list and want repeatable qualification before outreach.

Draft a reply to a LinkedIn prospect's message. Uses shared state, classification, and scripts for artifacts and metrics. Trigger on: "draft a reply", "they responded", or when the user pastes a prospect message.

Draft first-touch LinkedIn outreach for qualified leads using canonical lead artifacts. Uses persistent settings, brand voice, and outreach state to avoid duplicate sends. Trigger on: "draft message for [name]", "create outreach from filtered leads", or when a user selects leads from the lead-filter output.

First-run onboarding. Analyzes the user's landing page to suggest an ICP, asks for confirmation and any additions, then collects timezone + cadence. Saves everything to state/linkedin-settings.json and marks firstRunCompleted: true. Run this once before using any other skill.

Post-campaign analysis for LinkedIn outreach. Turns results into a precise diff of changes to three levers: voice, signals, and reply logic. Not a post-mortem, a fix list. Use when a campaign has finished and the user wants to know what to change. Trigger on: "review my campaign results", "my campaign just finished", "reply rate was low", "what do I fix after this campaign?", or any time the user shares metrics, message examples, or reply data and wants actionable next steps. Always use over generic analysis when outreach improvement is the goal.