Imported from MikeCheng1208/BattleTree (.agents/skills/ai-ark-similarity-search/SKILL.md). Install upstream with npx skills add MikeCheng1208/BattleTree --skill ai-ark-similarity-search. Copyright stays with the author.

Available Context

@_platform-references/org-variables.md

AI Ark Similarity Search

Goal

Find companies similar to a seed company using AI Ark's AI-powered similarity matching engine. This uses the lookalikeDomains parameter — AI Ark's unique differentiator with no Apollo equivalent.

Credit Cost

Each similarity search costs ~2.5 credits (uses the company search endpoint with lookalikeDomains). Use preview_mode: true to fetch 5 results first (same cost) before running a full search. Always warn the user before executing.

Required Capabilities

• AI Ark API: Similarity search via ai-ark-search edge function (action: 'company_search' with lookalike_domains)

Inputs

• seed_company_domain: Domain of the target company (preferred). Up to 5 domains for multi-seed matching. Examples: 'stripe.com', 'notion.so', 'hubspot.com', 'figma.com', 'salesforce.com'
• seed_company_name: Company name fallback if domain unknown
• match_count: How many lookalikes to find (default: 25, max: 100)
• preview_mode: Fetch 5 results first for quick validation

Execution

1. Identify the seed company domain (ask user if unclear)
2. Warn the user: "This search will cost ~2.5 AI Ark credits. Proceed?" (or offer preview first)
3. Optionally run with preview_mode: true to show 5 results
4. On confirmation, call ai-ark-search with seed domain(s)
5. Present results as a table for review

Output Contract

Return a table with columns:

• Company Name, Domain, Industry, Employee Count, Employee Range, Location, Description, Technologies, LinkedIn, Website

Show total_count to indicate how many lookalikes exist in the database.

Key Differentiator

Uses AI Ark's lookalikeDomains parameter on the company search endpoint. Accepts up to 5 seed domains or LinkedIn URLs. AI Ark matches by firmographic profile similarity (industry, size, tech stack, growth signals) — not keyword matching.

Common Patterns

• Seed from best customer -> find 100 lookalikes -> people search for VPs -> outbound
• Seed from closed-won deal -> lookalikes -> pipeline expansion
• Multi-seed: pass 3-5 top accounts as seeds -> find companies matching that ICP profile