Skip to content

skills.md

Latest commit

 

History

History
207 lines (151 loc) · 8.88 KB

skills.md

File metadata and controls

207 lines (151 loc) · 8.88 KB
Raw

Skills Catalog

Tip: you can run any of these commands by asking Claude Code in your IDE — see Quick start in README.

Skills are the execution primitives of GTM-OS. Each skill is a self-contained operation that can be invoked by the CLI, the orchestrator, or chained together in multi-step workflows.

Skills are provider-agnostic — find-companies doesn't know about Crustdata or Firecrawl. It declares what capabilities it needs, and the Provider Registry resolves the best available provider.

Quick Reference

Skill Category What it does Key provider
find-companies research Search companies by criteria Crustdata
find-people research Search people by title, company, role Crustdata
find-linkedin research Resolve LinkedIn URLs from name + email Crustdata
scrape-linkedin research Scrape post likers/commenters Unipile
competitive-intel research Research competitors (web + data) Firecrawl + Crustdata
research research Open-ended multi-step web research Anthropic + Firecrawl
qualify-leads analysis 7-gate ICP qualification scoring Anthropic
track-campaign outreach Poll providers, advance sequences, sync state Unipile + Instantly
enrich-leads data Add contact info, tech stack, signals Crustdata, FullEnrich
export-data data Export results as CSV/JSON None (local)
personalize content AI-personalize outreach messages Anthropic + Unipile
email-sequence content Generate email drip sequences Anthropic
answer-comments outreach Reply to LinkedIn post comments Unipile
reply-to-comments outreach Template-based LinkedIn replies Unipile
send-email-sequence outreach Send emails via Instantly Instantly
multi-channel-campaign outreach LinkedIn + email combined Unipile + Instantly
orchestrate integration Natural language → multi-step workflow Anthropic
visualize-campaigns analysis Campaign dashboard in browser None (local)
monthly-campaign-report analysis Cross-campaign intelligence report Anthropic
optimize-skill analysis RL-based skill tuning via swipe UI Anthropic

Detailed Descriptions

Research

find-companies

Search for companies matching specific criteria — industry, size, location, funding stage.

yalc-gtm orchestrate "find 10 SaaS companies in Berlin with 50-200 employees"

Inputs: query (natural language), count (default: 10) Providers: Uses Crustdata for structured search, Firecrawl for web search fallback. Output: Company list with name, website, industry, headcount, location.

find-people

Search for people by title, company, seniority level, and location.

yalc-gtm orchestrate "find VP of Engineering at Stripe"

Inputs: companies (array), titles (array), seniorityLevels (array), location, limit Providers: Crustdata people search. Output: People list with name, title, company, LinkedIn URL.

scrape-linkedin

Scrape likers and commenters from a LinkedIn post URL. Optionally auto-qualify results.

yalc-gtm leads:scrape-post --url "https://linkedin.com/feed/update/..."

Inputs: url, type (likers, commenters, both), maxPages, autoQualify Providers: Unipile. Output: People list with LinkedIn profile data.

competitive-intel

Research a competitor — scrape their website, pull company data, analyze positioning.

yalc-gtm orchestrate "research Gainsight's positioning and pricing"

Inputs: competitor (URL or name), enrichWithCrustdata (boolean) Providers: Firecrawl (web scraping) + Crustdata (company data). Output: Competitor profile with positioning, pricing, strengths, weaknesses.

Analysis

qualify-leads

Score and qualify leads against your ICP framework. Each lead gets a 0-100 score with reasoning.

Inputs: resultSetId, segment (optional — qualify against specific ICP segment) Providers: Anthropic (AI scoring against framework + intelligence). Output: Qualified leads with scores and reasons.

The 7-gate pipeline:

  1. Title match against ICP roles
  2. Company size check
  3. Industry alignment
  4. Disqualifier screening
  5. Signal scoring (buying intent signals, trigger events)
  6. Framework fit (positioning alignment)
  7. Final composite score

visualize-campaigns

Launch a visual dashboard showing campaign status, per-lead timelines, and variant performance.

Inputs: campaignId (optional), status filter, port (default: 3847) Output: Opens browser dashboard at http://localhost:3847

monthly-campaign-report

Generate a cross-campaign monthly intelligence report analyzing performance, variant winners, and strategic recommendations.

Inputs: month (YYYY-MM format), campaignIds (optional) Output: Structured report with metrics, insights, and next-month recommendations.

optimize-skill

Reinforcement learning for skills. Generates sample outputs, presents them in a swipe UI, and tunes prompts based on your preferences.

Inputs: skillId, samples (array of sample inputs), outputType, port Output: Opens swipe UI at localhost. Swipe right = like, left = dislike. Saves tuned prompt.

Data

enrich-leads

Enrich leads with additional data — contact info, tech stack, company signals.

Inputs: resultSetId, type (contact, company, tech-stack, full) Providers: Crustdata (company), FullEnrich (contact), Orthogonal (fallback).

export-data

Export a result set as CSV or JSON. No external provider needed.

Inputs: resultSetId, format (csv, json) Output: File written to ./data/ directory.

Content

personalize

AI-personalize an outreach message for a specific lead. Pulls LinkedIn profile, company signals, and intelligence to write a unique message.

Inputs: lead (object with email, name, company, linkedinUrl), template, enrichWithCrustdata, segmentId Providers: Anthropic + Unipile (LinkedIn profile) + Crustdata (company data). Output: Personalized message text.

email-sequence

Generate a complete email drip sequence based on your framework, ICP segment, and product context.

Inputs: type (cold-outreach, nurture, re-engagement, onboarding), segmentId, productContext, audienceContext Providers: Anthropic. Output: Multi-step email sequence with subjects, bodies, and timing.

Outreach

answer-comments

Reply to comments on a LinkedIn post. Modes: conversational (AI-generated replies) or lead-magnet (template with personalization).

Inputs: url, mode (conversational, lead-magnet), template, maxReplies, dryRun Providers: Unipile.

reply-to-comments

Template-based LinkedIn comment replies with personalization tokens. Lower-level than answer-comments.

Inputs: url, template (use {{name}} for first name), excludeAuthors, maxReplies, dryRun Providers: Unipile.

send-email-sequence

Send an email sequence via Instantly. Creates the campaign, adds leads, and launches.

Inputs: campaignName, leads (array), sequences (array of steps), dryRun Providers: Instantly. Prerequisite: At least one email account configured in Instantly.

multi-channel-campaign

Combined LinkedIn + email campaign. Sends LinkedIn connections via Unipile and email sequences via Instantly.

Inputs: sequencePath (YAML), leads, linkedinAccountId, dryRun Providers: Unipile + Instantly.

Integration

orchestrate

The meta-skill. Takes a natural language request, decomposes it into a multi-step workflow, selects skills and providers, and executes.

yalc-gtm orchestrate "find 10 companies matching my ICP, find their decision makers, qualify them, and create a LinkedIn campaign"

Inputs: query (natural language), autoApprove (skip confirmation gates) Providers: Anthropic (planning) + whatever the sub-skills need.

This is the most powerful skill — it chains other skills together based on your request and available providers.

Decision Tree

"I want to..."

  • Find companies → find-companies or orchestrate
  • Find people at a company → find-people
  • Get emails/phones for leads → enrich-leads (type: contact)
  • Score leads against my ICP → qualify-leads
  • Scrape a LinkedIn post → scrape-linkedin
  • Research a competitor → competitive-intel
  • Write personalized outreach → personalize
  • Generate email sequences → email-sequence
  • Reply to LinkedIn comments → answer-comments
  • Send cold emails → send-email-sequence
  • Run LinkedIn + email together → multi-channel-campaign
  • Do something complex → orchestrate (describe it in plain English)
  • See campaign performance → visualize-campaigns
  • Monthly report → monthly-campaign-report
  • Export data → export-data