Built for B2B founders, sales teams, and growth operators who do outbound on LinkedIn and want a repeatable, signal-driven system — without paying for a bloated sales tool.

What this does

You give it a CSV of LinkedIn leads. It filters, scores, drafts first-touch messages in your voice, handles replies, and tells you what to fix after each campaign.

Everything runs as Claude Code Skills — your AI handles the filtering, drafting, and reply logic. Bring your own leads, or use linkednav.com to source warm leads from your LinkedIn activity.

/onboard           → analyze your product, set your ICP
/calibrate-voice   → build your writing style guide
/linkedin-lead-filter → score and bucket your leads
/draft-outreach    → write first-touch messages, signal-first
/reply-handler     → classify and draft replies
/campaign-retro    → diagnose what to fix after a campaign

Who this is for

B2B founders doing outbound themselves
Sales reps and SDRs who want better personalization without more tools
Growth operators running LinkedIn campaigns at small-to-medium scale
Anyone who has tried LinkedIn outreach and found their reply rates too low

Requirements

Claude Code — the only hard requirement
Node.js 18+ — optional. If installed, file processing is faster. If not, Claude handles everything directly.

Quickstart

git clone https://github.com/xx254/linkedin_skills
cd linkedin_skills
bash setup

Then open the folder in Claude Code and run:

/onboard

That's it. /onboard analyzes your product, suggests your ICP, and saves your settings. All other skills are ready to use after that.

Skills — recommended order

/onboard → /calibrate-voice → /linkedin-lead-filter → /draft-outreach → /reply-handler → /campaign-retro

Every skill can be run independently. The only hard dependency: /linkedin-lead-filter requires /onboard to have been completed first. Everything else just works better with more context — it's not a blocker.

Skill	What it does	Input	Output	Run alone?
`/onboard`	Analyzes your landing page, suggests your ICP, asks you to confirm or adjust, then saves your settings. Run once.	Your landing page URL	`state/linkedin-settings.json`, `context/about-me.md`	✅ Start here
`/calibrate-voice`	Builds a writing style guide from your real messages (or a preset). Every draft skill follows this guide. Re-run when reply rates drop or your style shifts.	Paste real messages you've written, or pick a style archetype	`context/brand-voice.md`	✅ Only needs your messages or a style choice
`/linkedin-lead-filter`	Scores and buckets a CSV of LinkedIn leads against your ICP. Produces qualified/disqualified files and a canonical lead registry for downstream use.	CSV file path or pasted CSV content	`output/leads_<date>/` — categorized CSVs + `lead_registry_<date>.json`	✅ Requires `/onboard` first
`/draft-outreach`	Drafts a first-touch LinkedIn connection request for a lead. Signal-first, no pitch. Can take a lead from the filter registry or a name/company/title you type directly.	Registry path + lead key, or name / company / title / LinkedIn URL directly	`output/outreach_<date>/draft_<lead>_<date>.json` — draft + variant	✅ No prior filter run needed
`/reply-handler`	Classifies a prospect's reply (interested, hesitant, not interested, etc.) and drafts a contextual response. One question per message, no repositioning as a pitch.	Paste the prospect's message	Reply draft + classification + next-step strategy	✅ Most standalone skill
`/campaign-retro`	Diagnoses which lever is broken (targeting, voice, or reply handling) and outputs a precise fix list. Updates `brand-voice.md` with approved changes.	Your send/reply/acceptance metrics, or examples of messages that did/didn't work	Diagnosis report + edits to `context/brand-voice.md`	✅ Just paste your numbers

What you provide · What you get (overview)

What you provide	What you get
ICP / filter rules (saved to `linkedin-settings.json` via `/onboard`) + real lead CSV	Daily output: `qualified` / various disqualify CSVs + *`lead_registry_.json`** (the canonical key source for all downstream steps)
Voice calibration (saved to `brand-voice.md` via `/calibrate-voice`); optional: real messages you've written	First-touch LinkedIn drafts (written to `output/outreach_*` by script, generated by Agent per skill flow)
`lead_key` + prospect's message	Reply draft (`output/replies_*`)
`lead_key` + conversation outcome	Status and metrics update (recorded in `linkedin-system-state.json` for tracking)
Real data and examples during retro (shared in conversation with Agent) + optional weekly report script	Campaign retro conclusions + *`reports/weekly-summary_.md`**

Getting your leads

This pipeline works with any CSV of LinkedIn leads. You have two paths:

Option A — Bring your own CSV

Export from LinkedIn Sales Navigator, Apollo, or any other tool. The script accepts any CSV that can be mapped to name, title, company, and LinkedIn URL. See CSV column names below.

Option B — Use linkednav.com warm leads (recommended)

linkednav.com surfaces people who have already shown interest in you or your content — these are warmer than cold lists and respond at higher rates. Export your leads directly from the dashboard into this pipeline.

Signals LinkedNav tracks:

Category	Signal	What it means
Your content	`post_like`	Liked one of your posts
	`post_comment`	Commented on one of your posts
	`post_share`	Shared your content
	`poll_vote`	Voted on your poll
	`article_reaction`	Reacted to your article
	`article_read`	Read your long-form article
	`newsletter_subscriber`	Subscribed to your newsletter
	`mention`	Mentioned you in a post
	`content_download`	Downloaded your lead magnet
Profile & network	`profile_view`	Viewed your profile
	`connection_request`	Sent you a connection request
	`event_attendee`	Attended your LinkedIn event
	`company_follower`	Follows your company page
Competitor signals	`competitor_post_like`	Liked a competitor's post
	`competitor_post_comment`	Commented on a competitor's post
	`competitor_event_attendee`	Attended a competitor's event
Intent signals	`tool_recommendation_ask`	Posted asking for tool recommendations
	`pain_point_post`	Posted about a problem your service solves
	`influencer_post_like`	Engaging with niche influencers in your category
	`industry_event_attendee`	Attended a relevant industry event
	`conference_speaker`	Speaking at an industry conference
Company growth	`funding_news`	Company announced funding
	`press_mention`	Company featured in the press
	`hiring_surge`	Company posted many roles recently
	`hiring_adjacent_role`	Hiring for a role that typically buys your service
	`hiring_pain_point_jd`	Job description describes the problem you solve
	`hiring_competitor_tool_mention`	JD requires experience with a competing tool
	`hiring_volume_surge`	Function expanding fast
	`hiring_new_senior_leader`	New C-suite hire — tool stack reassessment likely
Life events	`job_change`	Recently changed roles
	`work_anniversary`	Work anniversary — often reassessing tools

All signals map directly to the scoring logic in /linkedin-lead-filter. High-warmth signals score 3 and are prioritized for outreach. Reaching out within 24 hours of a signal → significantly higher reply rates.

LinkedNav also handles sending. After you've drafted and approved your messages here, LinkedNav can run the outreach campaign and follow-ups autonomously.

What you need to provide (user input / configuration)

Each item below is a real data entry point. Do not use fabricated placeholder data in place of real leads. If you don't have data yet, complete configuration first, then import a real CSV.

CSV column names

Your CSV needs to be mappable to name, title, company, and LinkedIn URL at minimum. The skill accepts common export formats from LinkedIn Sales Navigator, Apollo, linkednav.com, and others.

Supported column aliases:

Name: First Name + Last Name, or Name
Title: Job Title / Position / Title
Company: Company
Industry: Industry
Location: Location / Geography
LinkedIn: LinkedIn URL / Linkedin URL / LinkedIn
Signal: Source Post, Imported At

See sample_contacts.example.csv for a working example.

For full technical details see PRODUCT_SYSTEM.md and contracts/lead-artifact-contract.md.

Distribution notes

output/ and reports/ contain run artifacts and are excluded via .gitignore; directories are created by bootstrap-system.js.
state/*.json is excluded from the repo — only state/*.example.json templates are committed. Copy them or run bootstrap after cloning.
*.csv files are excluded from the repo — never commit real lead data. Use sample_contacts.example.csv as a structural reference.
After cloning, run bootstrap first, then fill in your own CSV and context files.

What a warm lead CSV looks like

This is the kind of data you get from linkednav.com. Each row is a real person who already showed a signal — not a cold name from a list.

Name	Title	Company	Signal	What happened
Jane Doe	VP of Marketing	Acme Corp	`post_like`	Liked your post on outbound sequencing
John Smith	Head of Growth	Beta Inc	`post_comment`	Commented: "this is exactly the problem we're dealing with"
Maria Garcia	Director of Demand Gen	Gamma Ltd	`profile_view`	Viewed your profile 3 times in 7 days
David Lee	Founder	Delta Ventures	`post_share`	Shared your post on LinkedIn automation to their network
Sarah Kim	CMO	Epsilon Co	`newsletter_subscriber`	Subscribed to your LinkedIn newsletter on GTM strategy
Alex Johnson	VP Sales	Zeta Corp	`job_change`	Started new role as VP Sales — likely evaluating new tools
Priya Patel	Growth Lead	Eta Startup	`event_attendee`	Attended your LinkedIn event on warm outbound
Tom Chen	Marketing Director	Theta Inc	`company_follower`	Follows your company page
Lisa Wang	Revenue Operations Lead	Iota Corp	`poll_vote`	Voted on your poll about outreach personalization
Marcus Brown	Head of Partnerships	Kappa Inc	`article_reaction`	Reacted to your article on LinkedIn signal tracking
Nina Roberts	Sales Director	Lambda Co	`connection_request`	Sent you a connection request without a note
James Park	CEO	Mu Ventures	`mention`	Mentioned you in a post about outreach tools
Rachel Chen	VP of Sales	Nu Corp	`competitor_post_like`	Liked a competitor post within the last 24 hours
Kevin Morris	Head of Revenue	Xi Inc	`competitor_post_comment`	Commented on a competitor post: "we've been looking for exactly this"
Aisha Okafor	Growth Marketing Manager	Omicron Ltd	`influencer_post_like`	Liked 4 posts from a niche influencer in your category this week
Wei Zhang	CFO	Pi Ventures	`funding_news`	Company announced $12M Series A — now scaling GTM team
Elena Sousa	Director of Operations	Rho Software	`hiring_surge`	Company posted 14 new roles in the last 30 days
Daniel Osei	Co-Founder	Sigma Labs	`tool_recommendation_ask`	Posted asking for tool recommendations for LinkedIn outreach
Yuki Tanaka	Marketing Ops Manager	Tau Corp	`pain_point_post`	Posted about struggling with low reply rates on outbound
Sofia Rossi	Chief of Staff	Upsilon Co	`article_read`	Read your long-form article on signal-based outreach
Ethan Brooks	VP Marketing	Phi Group	`content_download`	Downloaded your lead magnet on outreach templates
Amara Diallo	Head of Sales Enablement	Chi Inc	`work_anniversary`	5-year work anniversary — likely reassessing tools and processes
Lucas Andrade	Founder	Psi Startup	`press_mention`	Company featured in TechCrunch for new product launch
Hannah Müller	CRO	Omega GmbH	`conference_speaker`	Announced as keynote speaker at SaaStr 2024
Ryan Thompson	VP of Operations	Alpha SaaS	`hiring_adjacent_role`	Hiring for a role that typically owns the problem your service solves
Fatima Al-Rashid	Head of Marketing	Beta Cloud	`hiring_pain_point_jd`	Job description explicitly describes the pain your service solves
Liam O'Brien	Director of Operations	Gamma Tech	`hiring_volume_surge`	Same function posted 4 roles in 30 days — expanding fast
Mei Suzuki	CEO	Delta Co	`hiring_new_senior_leader`	Just hired a new C-suite leader in your buyer persona
Carlos Mendez	VP of Product	Epsilon Media	`hiring_competitor_tool_mention`	JD requires experience with a competing tool — actively evaluating
Ingrid Larsson	Head of Growth	Zeta Nordic	`competitor_event_attendee`	Attended a competitor's product launch event last week
Omar Hassan	VP of Partnerships	Eta Corp	`industry_event_attendee`	Attended industry conference on B2B sales automation