# Lessie Skills & CLI: Make Your AI Agent Read People and Markets ## TL;DR - Lessie Skills & CLI dissolves the SaaS wall, turning Lessie’s people-search and company-intelligence capabilities into liquid skills any MCP-compatible agent can call directly. - Install in seconds: `npx skills add LessieAI/lessie-skill -y -g` - Compatible with Claude Code, OpenClaw, Codex, and every agent environment that speaks MCP. - Three battle-tested workflows: B2B precision growth for an AI video tool, overseas KOL audit & outreach, and headhunting top AI researchers. - Backed by PeopleSearchBench — the industry’s first open benchmark for AI people search, with 119 reproducible queries. ## Our Intent: Tear Down the SaaS Wall, Make Lessie Liquid For a long time, traditional SaaS thinking has equated a product with a “web page” or an “app,” locking powerful capabilities behind beautiful but sealed dashboards. As deep work has migrated, at scale, into **Claude Code, Codex, and custom agent environments**, this leaves products stranded in an awkward gap: the capability is rich, but it cannot be _called_. That is exactly why, on top of Lessie’s existing web experience, we are rolling out **Skills & CLI** end-to-end. It does not mean Lessie is walking away from the web. On the contrary, we believe a truly AI-native product should be **liquid**: - **On the web**— it offers intuitive, visual feedback, well suited to humans making complex commercial decisions. - **In the CLI**— it becomes lightweight, frictionless skills that act as the agent’s “endogenous knowledge” and execution limbs. We are not building yet another tool. We are reshaping **how capability gets delivered**. In the previous SaaS era, humans were data “couriers,” exhausted by toggling between tabs. Inside Lessie’s logic, we refuse to let data sit at the level of “retrieved and summarized.” Through Skills & CLI, we encapsulate Lessie’s core capabilities — commercial scouting, people profiling, and company intelligence — so they are not just usable by humans, but **natively callable by agents**. We refuse to repeat the world inside an information stream. We deliver business truth that cuts straight to the endgame, through an evolving toolchain. When your agent learns to read people and markets through Lessie skills, it shatters the fourth wall of the chatbox — no longer a dialogue loop spinning in place, but a **digital partner with commercial intuition**, fighting alongside you. ## Quick Start: Two Commands, an Agent That Reads People and Markets Before you begin, you need one thing: a **Lessie account**. [Sign up at app.lessie.ai/find](https://app.lessie.ai/find) and trial credits are issued instantly. Lessie Skills are compatible with **Claude Code, OpenClaw, Codex**, and every agent environment supporting the MCP protocol. ### Step 1 — Pull the people-search and email skills locally Visit the [Lessie Skills GitHub repository](https://github.com/LessieAI/lessie-skill) and run this single command in your terminal: `npx skills add LessieAI/lessie-skill -y -g` This command pulls both the **people-search** skill and the **email** skill into your agent’s skill directory. Once installed, your Claude Code or Codex has just “learned” new actions: finding people, profiling companies, verifying contacts, and sending email. ### Step 2 — Verify install & initialize Restart your agent (Claude Code is shown as the example) so it loads the freshly installed skill. In the terminal, simply tell the agent `/people-search` or `/email`. The agent will automatically check CLI status and complete initialization. On first init, the system will pop up Lessie’s authorization page — just click **Authorize**. When the terminal prints `authorized: true`, you’re all set. ### Best Practices #### Scenario 1 — B2B Acquisition: Global Precision Growth for an AI Video Tool You are the CEO of an AI video generation tool (hyper-realistic digital humans, multilingual lip-sync, one-click ad creative). The pain point: you cannot find genuinely differentiated users, and traditional social DMs are routinely treated as spam —making it nearly impossible to reach the enterprise decision-makers with the highest appetite for video marketing. 1. Mine product differentiation and lock the persona: find the “business fit” angle `/people-search` My product is an AI video generation tool. Core advantage: multilingual lip-sync and ultra-fast rendering. Find prospects in North America that are currently building multilingual channels or cross-border e-commerce. Analyze which parts of their business (TikTok ads, overseas customer service video) my product can amplify. 2. Find the people and “see through” the org: penetrate to the “video budget owner” `/people-search` Inside the target accounts, lock decision-makers at the level of Head of Growth or Creative Lead. Run enrich-people to grab their corporate emails plus their recent LinkedIn updates about content innovation. 3. Send the email: a “cost-down, output-up” personalized pitch `/email` Combine the business angles surfaced in step one. For one of these contacts, generate and send a personalized cold email that lays out the concrete math of “80% lower cost / 10x more output.” #### Scenario 2 — KOL Partnerships: Precision Audit and Outreach for Overseas Creators You are the head of overseas growth for an AI product, hunting for international KOLs to amplify it. The pain: you do not know the overseas social ecosystem well, blind outreach is brutally inefficient, and although YouTube and Twitter are full of tech creators, follower counts are heavily inflated and manual screening drains the entire week. 1. Find the lookalikes and “mine the competitor”: lock high-conversion ground `/people-search` My product is X. Find the head competitor in the same lane and pull every creator they’ve worked with on YouTube, TikTok, and Twitter in one shot. Have the AI auto-cluster them and recommend 30 vertical creators with audience overlap above 80%. 2. Audit the creators and “dehydrate” the list: lock real influence `/people-search` Audit the creators from step one. Auto-eliminate any account whose bot follower ratio exceeds 40% or whose engagement signals are anomalous. 3. Mine their content and “custom-tailor the script”: invitations that match the creator’s voice `/people-search` Deeply analyze the last 10 review videos of one target creator: their logic, visual style, and language preference (hardcore evaluator vs. playful demoer). Auto-generate a partnership invite plus a draft video script that matches their persona and tone. #### Scenario 3 — Recruiting: Hunting Top AI Researchers You are the HRD of an Agent startup. You urgently need to hire AI researchers fluent in Harness engineering to architect a high-availability Agent stack. The pain: this kind of talent typically hides inside top labs or big tech, never opens job platforms, and their true “engineering landing ability” is impossible to read from a flat resume. Establishing initial trust is even harder. 1. Find the people and analyze: lock the “hardcore engineering” researchers `/people-search` Search GitHub contributors and ArXiv authors. Lock experts who, in the past year, have shipped a high-star project or core paper in the “Agent Evaluation Harness” space. 2. Get the contact: penetrate the academic and open-source private domain `/people-search` Enrich their personal email and recent X/social feeds. 3. Send the email: a “deep technical fit” targeted invitation `/email` Drawing on the candidate’s specific work in Harness engineering, write and send a deep invitation framed as a discussion about Agent architecture stability and evaluation standards. ## AI-Native Era: Reshaping “Cut-to-the-Endgame” Certainty In the AI age, being able to “search things up” is no longer a skill. Generating **actionable intuition** from data is. Lessie Skills exists to pull the agent out of its “bookworm” phase. When your agent learns to read people and markets, it stops being a tool and becomes a **Digital Partner** fighting alongside you. To make all of this verifiable, Lessie AI has released the industry’s first open benchmark for AI people search — **PeopleSearchBench**. It covers 119 real-world business queries across the four core scenarios of recruiting, B2B prospect finding, KOL discovery, and expert sourcing. It runs head-to-head against Exa, Claude Code, Juicebox, and other mainstream platforms. Every result is independently web-verified— no self-reported numbers, fully reproducible from end to end. See [lessie.ai/benchmark](/content/benchmark/index.html) for details. **Start now. Two commands. Wake up your agent’s commercial nature.** `npx skills add LessieAI/lessie-skill -y -g` ## FAQ ### Which AI agents are compatible with Lessie Skills? Lessie Skills works with [Claude Code, OpenClaw, Codex](/content/lessie-skills-cli/index.html), and every agent environment that supports the MCP (Model Context Protocol). If your agent speaks MCP, it can call Lessie Skills natively. ### How do I install Lessie Skills? Run `npx skills add LessieAI/lessie-skill -y -g` in your terminal. Restart your agent and type `/people-search` or `/email` to initialize. The agent handles authorization automatically; if the CLI binary is missing, it auto-installs the dependency. ### What can Skills & CLI do that the web dashboard cannot? Skills expose the same core capabilities as the web platform (people search, company intelligence, contact enrichment, email outreach), but as composable actions inside your agent’s reasoning loop. Your agent can chain multiple skills in a single workflow — search prospects, score intent, enrich contacts, send a personalized email — without you switching tabs. ### Is Lessie Skills free to try? Yes. [Sign up for a Lessie account](https://app.lessie.ai/find) and trial credits are issued instantly. Skills consume the same credits as web searches. ### How is Lessie’s search quality measured? Through [PeopleSearchBench](/content/benchmark/index.html), the industry’s first open benchmark for AI people search. 119 reproducible business queries span recruiting, B2B prospecting, KOL discovery, and expert sourcing, scored head-to-head against major platforms.