Lead generation

Claude skills for sales prospecting and outbound: what a skill can do, and when you need an MCP server

Claude skills for sales prospecting: what a skill can and can't do, when you need an MCP server, and how the traxy MCP finds leads and runs outreach.

Claude skills for sales prospecting and outbound: what a skill can do, and when you need an MCP server — traxy blog

TL;DR: Yes, there are Claude skills for sales prospecting and outbound. lemlist publishes a set of Claude skills for writing sequences, defining an ICP and handling replies, and Anthropic's plugin directory lists a Sales plugin and a Common Room plugin with prospecting and outreach skills. A skill is a set of instructions Claude follows, though, so finding live leads and sending LinkedIn or email messages need a connected tool such as an MCP server. The traxy MCP is that connection for LinkedIn GTM: it finds high intent leads from LinkedIn likes, posts and comments, researches each one, and runs LinkedIn and email campaigns from your own connected accounts, with your yes in the chat before anything sends.

We make traxy, so read this as the vendor's guide. It explains what a Claude skill is, what one can and can't do for prospecting, when you need a connected MCP server, and how the traxy MCP covers both halves of outbound. Third-party skills and plugins are described only from their own pages, linked in each section and checked November 2026. traxy's steps come from the traxy docs linked below.

Are there Claude skills for sales prospecting and outbound?

Yes. Several vendors and Anthropic itself publish skills for sales work, and you can write your own. The ones a sales team is most likely to meet, checked November 2026:

  • lemlist publishes "Claude skills for GTM engineering": its page says "38 skills across 6 categories", with skills such as People Finder, List Builder, Linkedin Sequence, Reply Handler, ICP Definer and Trigger Finder, and a download of all of them from GitHub (lemlist Claude skills).

  • Anthropic's Sales plugin, listed as Anthropic verified, covers account research, call prep, research-first outreach drafting and call summaries, and works with the CRM, email, calendar and chat tools your team already uses or with pasted notes and CSV uploads (Sales plugin).

  • The Common Room plugin in Anthropic's plugin directory lists Call Prep, Account & Contact Research, Compose Outreach, Prospecting and a Weekly Prep Brief, and says it will "Draft emails, call scripts, and LinkedIn messages" (Common Room plugin).

What a skill gives Claude is know-how: how to define an ICP, how to structure a LinkedIn sequence, how to answer an objection. Anthropic describes the split plainly: MCP "connects Claude to external services and data sources," while skills provide procedural knowledge about how to do a task, and the two can be used together (What are Skills?). So the buyer's real question, "can a skill find my leads and send my outreach?", has a short answer: not on its own. The leads and the sending come from the product behind a connection.

That is where traxy fits. traxy runs your entire LinkedIn GTM end to end, from finding high intent leads from social intent to enriching and researching each one, to running relevant, personalized LinkedIn and email outreach to them on autopilot. From Claude, that runs through the traxy MCP: once you approve a campaign it runs on its own, or you approve every message.

What is a Claude skill? Instructions and steps, not data or sending

A Claude skill is a folder with a SKILL.md file, plus optional scripts and reference files, that Claude loads when your request matches it. Anthropic's docs describe skills as packaging "instructions, metadata, and optional resources (scripts, templates) that Claude uses automatically when relevant" (Agent Skills overview).

How a skill loads, per the same page:

  1. Metadata, always loaded. The skill's name and description sit in Claude's context from the start. Claude matches your request against the description to decide whether to use it.

  2. Instructions, loaded when triggered. The body of SKILL.md, with the workflow and guidance, enters the context only when the skill fires.

  3. Resources and code, loaded as needed. Extra files such as a messaging framework or a scoring rubric are read only when referenced, and bundled scripts run with only their output entering context.

Three details from Anthropic's docs matter for sales teams:

  • Network access depends on where the skill runs. On the Claude API, skills run in a sandbox with no network access. In claude.ai, access can be full, partial or none depending on settings. In Claude Code, skills have the same network access as any program on your computer.

  • Custom skills don't sync across surfaces. A skill uploaded to claude.ai is not on the API, and Claude Code skills live on the file system, so each surface needs its own copy.

  • Use skills only from trusted sources. Anthropic warns that a malicious skill "can direct Claude to call tools or run code in ways that don't match the Skill's stated purpose," and says to audit every bundled file before use.

None of that gives Claude a list of people who engaged with a LinkedIn post this week, or access to your LinkedIn account. A skill tells Claude how to work. The data it works on, and any action in another product, have to come from somewhere else.

What can a skill alone do for prospecting, and what can't it do?

A skill alone can shape how Claude thinks and writes about prospecting. It can't supply the prospects or send anything to them. Here is the split for common outbound jobs:

Job

A skill alone

What it needs from a connection

Define an ICP

Walks Claude through a framework and asks the right questions

Your real leads, to check the ICP against who engages

Write a LinkedIn or email sequence

Applies your message rules, tone and structure

The lead's signal and research, so each message has a real reason

Score a list

Applies a rubric to rows you paste or upload

Fresh leads from a source, with the activity behind each one

Find who is in market now

Explains which signals to look for

A product that watches LinkedIn activity and qualifies the people

Research a lead

Lists what to check before you write

Tools that read the lead's profile, posts and company

Send the outreach

Nothing to send from

A connected outreach tool with your sender accounts

Handle replies

Suggests an answer to a reply you paste

The inbox, the sequence state and a way to send the answer

So a skill is a good place for your team's playbook: your ICP notes, the claims you never make, the message length you like, the order you research things in. It is the wrong place to expect a lead source. If a skill's page says it finds contacts, read what it connects to. lemlist's own descriptions say People Finder works with "lemlist's people database" and List Builder uses "lemlist's search and signals", so the data comes from lemlist, not from the skill file.

When do you need a connected MCP server? Lead data, research and sending

You need a connected MCP server as soon as Claude has to read data that lives in a product or change something in it. For prospecting and outbound that means three jobs:

  • Lead data. Who engaged, who matches your ICP, which post or comment they left. Claude can't see LinkedIn engagement by itself.

  • Research on live data. Reading a lead's recent posts, company and the facts your campaign needs, then keeping that research on the lead.

  • Sending and follow-up. Launching a sequence from your own LinkedIn and email accounts, stopping it when someone replies, and answering in the inbox.

Before you connect any server, check three things: what it can read, what it can change or send, and whether it asks before acting. Our guide to MCP servers for sales compares servers by job: lead finding, contact data, CRM and LinkedIn.

Skills and MCP servers are not rivals. A skill that holds your team's playbook and an MCP server that holds the data and the actions work well side by side, which is the pattern Anthropic describes for partner skills built to work with their own connectors.

How does the traxy MCP cover both halves: finding high intent leads and running LinkedIn and email campaigns?

The traxy MCP lets Claude find and research high intent leads from LinkedIn activity, then draft, test and launch LinkedIn and email campaigns to them from your own connected accounts, with your yes in the chat before anything sends or launches. It also works in ChatGPT, Cursor, Codex and Grok. Step by step:

Step 1: connect traxy to Claude

In traxy, open Integrations > Claude and copy the MCP URL, https://api.traxy.ai/api/v1/mcp. Add it as a remote custom connector in Claude's connector settings, sign in to traxy when prompted, pick a default workspace and any others, review the requested actions and approve. No traxy REST API key is needed (Claude setup docs).

The public tool catalog lists 234 tools before scope and workspace filtering. What your connection can call depends on its scopes and your role, so Claude sees the tools your connection allows, and role, product access and action checks still apply (MCP tools reference).

Step 2: set your ICP, signals and tracking from chat

Ask Claude to read your ICP and Signal Brief and propose changes. traxy's docs say to read the current version before changing it, and arrays you supply replace the old values, so ask Claude to show the change first. You can also add tracked people, single posts by URL and company pages by domain, and traxy qualifies whoever engages with them.

Step 3: find high intent leads from social intent

This is the first half. From Claude you can:

  • Search social intent in plain English across 71M+ scanned LinkedIn likes, posts and comments, describing the person and the need.

  • List your qualified leads. A person counts as a lead only with a 50%+ match to your ICP, each lead is labeled Warm, Related or Activity, and every lead shows the post or comment behind it.

  • Start a discovery run or qualify the audience of one post. Both ask you to confirm first, because they use capacity.

  • List leads from every source traxy watches, including people who engage with competitors' and creators' posts and, if you turn on Profile views in Targeting, your profile viewers.

For prompts and a worked session, see how to find LinkedIn leads in Claude with the traxy MCP.

Step 4: research each lead

Ask Claude to research a lead or a group of leads for the facts you care about, such as what they posted, their role and their company. Signals outside LinkedIn, such as hiring or funding, come from that research on each lead, not from a live event feed. From Claude, a contact lookup runs only when you ask for it. If you turn on Auto-Enrich, traxy looks up contact data for newly qualified leads.

Step 5: draft, test and launch a LinkedIn and email campaign

This is the second half. Ask Claude to build a campaign for a set of leads. traxy researches each lead, groups leads by signal and prepares a personalized message variant for each group, as LinkedIn and email steps in one sequence. Then:

  1. Test on sample leads. Nothing is sent. You check whether the research supports each claim and look at skipped results as well as the drafts.

  2. Review the launch. See who will join, who is skipped (leads held by another unfinished campaign, and people a completed campaign already contacted), the sender, the steps and the delivery mode (review and launch docs).

  3. Say yes in the chat. traxy's docs say to preview campaign launches and message sends before confirming, and that "a draft or preview doesn't authorize execution."

  4. Let it run, or approve each message. After you approve the launch, the campaign runs on autopilot from your own connected LinkedIn and email accounts. If you prefer, set it so every message waits in a review queue for your approval.

Many campaigns can run at once on different signals, each with its own leads and messages. How the outreach itself works is on LinkedIn outbound with traxy.

Step 6: answer replies from the same chat

Replies, unsubscribes and hard bounces stop the sequence for that lead. From Claude you can read inbox conversations and send a reply after your yes. You can also pause, resume or stop a lead's sequence and manage do-not-contact entries. The full list of what the connection does is in what the traxy MCP server can do, and the whole flow end to end is in the LinkedIn GTM MCP workflow.

Third-party skills for sales, described from their own pages

traxy wrote this section and sells the traxy MCP. Each skill or plugin below is described only from its own page, linked in the row and checked November 2026. These pages change often, so confirm details with each vendor.

Skill or plugin

What its page says it does

Where it fits better than traxy

lemlist Claude skills (page)

"38 skills across 6 categories" (Writing, Analysing, Extrapolating, Building, Calling and Moderating). Examples: Linkedin Sequence "Writes a 2-message LinkedIn DM sequence after a connection is accepted"; Reply Handler crafts a response to a cold outreach reply; ICP Definer defines and prioritizes ICPs; People Finder works with lemlist's people database; List Builder uses lemlist's search and signals

Your team runs outbound in lemlist and wants Claude to follow ready-made playbooks for copy, ICPs and reply handling

Anthropic Sales plugin (page)

Listed as Anthropic verified, with its source in Anthropic's knowledge-work-plugins repository. Commands for call summaries and forecasts, and skills for account research, call prep, daily briefings and research-first outreach drafting. Works with the CRM, email, calendar and chat tools your team already uses, or standalone with pasted notes or CSV uploads. It "drafts or sends email" when you ask

Account executives who want call prep, follow-ups and forecasts on top of their CRM and email, rather than a source of new LinkedIn leads

Common Room plugin (page)

"Turn Common Room into your GTM copilot." Account and contact research, call briefs, prospect lists and account plans based on signal data, and drafts of emails, call scripts and LinkedIn messages. The page doesn't say it sends them

Teams that already use Common Room for signals and want Claude to research accounts and draft outreach from that data

traxy MCP (the LinkedIn GTM MCP)

Not a skill: a remote MCP connection. Finds high intent leads from LinkedIn likes, posts and comments with a 50%+ ICP match and the activity behind each one, researches them, and drafts, tests, launches and manages LinkedIn and email campaigns and inbox replies from your connected accounts, after your yes in the chat

You want Claude to bring the leads and run the LinkedIn and email outreach to them from one connection

Where others fit better, in short: if you only need Claude to write better copy or follow a sales method, a skill such as lemlist's, downloadable from GitHub, is enough. If your day is calls, follow-ups and forecasts in a CRM, Anthropic's Sales plugin is built for that. If your signals already live in Common Room, its plugin works from them. traxy fits when the hard part is knowing who on LinkedIn is in market now and reaching them over LinkedIn and email without moving lists between tools.

How do you choose between a skill, an MCP server, or both?

  • Start with the question you ask Claude most. "How should I write this?" points to a skill. "Who should I write to, and send it" points to an MCP server.

  • Check where the data comes from. If a skill mentions a database or signals, find the product behind them and what account it needs.

  • Check who sends, and from which account. With traxy, messages go from your own connected LinkedIn and email accounts, and nothing sends until you approve.

  • Check the approval step. Look for a preview and an explicit yes before a launch or a send, and a way to approve each message if you want that.

  • Audit any skill you install. Read the SKILL.md and scripts, as Anthropic advises, especially in Claude Code where skills have network access.

  • Combine them when it helps. Keep your playbook in a skill and the leads and actions in the MCP server.

For prompts that work across both halves, see Claude and ChatGPT prompts for LinkedIn GTM.

What are the limits?

  • The tools Claude can call depend on your connection's scopes, your role and the workspace. Seeing a tool in the catalog doesn't mean every workspace can use it.

  • Sending needs connected sender accounts. After approval, delivery can still wait on sender availability, schedule and limits, so check each lead's status.

  • traxy doesn't promise zero LinkedIn restriction risk. Keep volumes sensible for each sender account.

  • An ICP match is a fit score, not a chance that someone buys. Engagement on a post shows interest, not a decision.

  • Signals beyond LinkedIn activity come from research on each lead, not a live event feed.

  • A queued delivery or lookup doesn't prove it finished. Ask Claude to check the result before retrying.

FAQ

Can a Claude skill find leads on LinkedIn?

Not on its own. A skill gives Claude instructions, and Claude can't see who engaged with a LinkedIn post. With the traxy MCP connected, Claude can search 71M+ scanned LinkedIn likes, posts and comments and list your qualified leads with the activity behind each one.

Can a Claude skill send LinkedIn messages or emails?

A skill has no LinkedIn or email account to send from. Sending needs a connected outreach tool. Through the traxy MCP, Claude can send a one-off message or launch a LinkedIn and email campaign from your own connected accounts after you confirm in the chat.

Is a Claude skill the same as an MCP server?

No. Anthropic describes skills as procedural knowledge and MCP as the connection to external services and data. A skill tells Claude how to do a job; an MCP server gives Claude the data and the actions in a product.

Does anything send without my approval through the traxy MCP?

No. Claude shows what it will send or launch and waits for your yes, and traxy's docs say a draft or preview doesn't authorize execution. Once you approve a campaign it runs on autopilot, or you can approve every message in the review queue.

Which AI clients work with the traxy MCP?

Claude, ChatGPT, Cursor, Codex and Grok. In Claude you add https://api.traxy.ai/api/v1/mcp as a custom connector and sign in with OAuth, with no REST API key.

Are third-party sales skills safe to install?

Anthropic says to use skills only from trusted sources, such as ones you wrote or got from Anthropic, and to audit every file in a skill from anywhere else. A skill can direct Claude to call tools or run code, so treat installing one like installing software.

Back to all articles
The right people are already out there.

Find your warmest leads.
Send them personalized messages.

From relevant interest to personalized outreach.
Put traxy to work.

7-day free trial