icypeas-automation

Overview of icypeas-automation skill

What icypeas-automation does

icypeas-automation is a Claude skill for running Icypeas lead research and enrichment tasks through Composio’s Rube MCP. Instead of hard-coding one fixed Icypeas action, the skill teaches the agent to discover the current Icypeas tool schema first, verify the connection, and then execute the right Rube tool for the user’s sales or research workflow.

Best fit for lead research teams

This skill is most useful when you already use Icypeas or want an AI agent to help with prospect discovery, email finding, company/person enrichment, or related lead research operations. It fits sales operations, growth teams, recruiting researchers, agency prospecting workflows, and technical users who want Claude to call Icypeas tools safely instead of only drafting manual research instructions.

Key differentiator: schema discovery first

The most important behavior in the icypeas-automation skill is its insistence on calling RUBE_SEARCH_TOOLS before execution. That matters because MCP tool schemas can change, and Icypeas actions may expose different required fields depending on the current Composio toolkit. This reduces failed calls, stale examples, and hallucinated parameters compared with a generic prompt.

Adoption requirements and limits

To use this skill, Rube MCP must be available in your AI client, and the Icypeas toolkit connection must be active through RUBE_MANAGE_CONNECTIONS. The repository for this skill is intentionally lightweight: the main guidance is in SKILL.md, with no extra scripts, rules, or reference files. If you need a complete CRM sync pipeline, validation layer, or custom deduplication logic, you will likely need to add that around the skill.

How to Use icypeas-automation skill

icypeas-automation install and setup path

Install the skill from the source repository with:

npx skills add ComposioHQ/awesome-claude-skills --skill icypeas-automation

Then configure Rube MCP in your client by adding https://rube.app/mcp as an MCP server. The skill depends on the rube MCP server, not a local script. Before asking Claude to run an Icypeas workflow, confirm that RUBE_SEARCH_TOOLS is available and use RUBE_MANAGE_CONNECTIONS with toolkit icypeas to check whether the connection is ACTIVE. If it is not active, complete the returned authorization flow first.

Inputs the skill needs for reliable execution

A strong icypeas-automation usage prompt should include the business objective, lead criteria, required output fields, volume, acceptable uncertainty, and any exclusion rules. For example, “Find likely sales operations leaders at B2B SaaS companies in Germany” is better than “find leads” because the agent can search for the right Icypeas tools and map your criteria into the live schema.

Useful inputs include:

• Target persona, seniority, geography, industry, company size, or domain list
• Whether you need people, companies, emails, enrichment, or verification
• Output format such as CSV-style table, JSON, CRM-ready fields, or review summary
• Constraints such as “do not enrich personal emails” or “skip existing accounts”
• Success criteria such as minimum confidence, maximum records, or required fields

Turning a rough goal into a complete prompt

A weak prompt says: “Use Icypeas to get leads.”

A stronger prompt says: “Use the icypeas-automation skill for Lead Research. First discover the current Icypeas tools with RUBE_SEARCH_TOOLS, then confirm the Icypeas connection is active. I need up to 50 prospects: VP Sales, Head of Revenue, or Sales Operations leaders at B2B SaaS companies with 50–500 employees in France and Belgium. Return name, title, company, company domain, LinkedIn URL if available, business email if available, source tool used, and confidence notes. Do not include agencies or consultancies.”

That prompt improves results because it gives the agent enough context to choose tools, populate required fields, and judge whether returned records match the research goal.

Files to read before using the repository

Start with composio-skills/icypeas-automation/SKILL.md. It contains the actual operational pattern: prerequisites, setup, tool discovery, connection checking, workflow execution, and error-handling expectations. There are no additional README.md, resources/, references/, rules/, or scripts/ files in the skill folder, so do not expect hidden implementation logic. The practical value is in the MCP workflow discipline, especially tool discovery before execution.

icypeas-automation skill FAQ

Is icypeas-automation only for Icypeas users?

Yes, in practice. The skill is designed around Composio’s Icypeas toolkit via Rube MCP. You can install it without an active Icypeas connection, but it will not execute useful lead research workflows until the toolkit connection is authorized and active.

Why not just ask Claude to research leads manually?

A generic prompt can draft research criteria or suggest search strategies, but it cannot reliably call current Icypeas actions unless the MCP tools are available and discovered. The icypeas-automation skill adds a repeatable pattern: search current tool schemas, check authentication, execute with known fields, and handle tool-level failures.

Is this beginner-friendly?

It is beginner-friendly if your AI client already supports MCP and you are comfortable authorizing external tool connections. It is less suitable for users who expect a one-click spreadsheet app. The skill assumes you can verify Rube MCP availability, inspect tool responses, and clarify lead criteria when the agent asks follow-up questions.

When should I not use this skill?

Do not use icypeas-automation when you need unsupported data sources, fully offline research, guaranteed email accuracy without verification, or complex downstream CRM governance. It is also a poor fit if you cannot share the necessary prospecting criteria with the agent or if your organization requires manual approval before every third-party enrichment call.

How to Improve icypeas-automation skill

Improve icypeas-automation results with tighter criteria

The best way to improve icypeas-automation output is to narrow the research request before execution. Replace broad categories like “tech companies” with explicit filters: “B2B SaaS companies selling HR software, 100–1,000 employees, headquartered in the UK, excluding staffing agencies.” Clear filters reduce irrelevant enrichment calls and make it easier to judge whether the returned leads are usable.

Common failure modes to watch for

Most failures come from inactive connections, skipped schema discovery, missing required fields, or vague lead definitions. If a tool call fails, ask the agent to re-run RUBE_SEARCH_TOOLS for the specific use case and compare the attempted parameters against the returned schema. If results are low quality, refine the ICP before increasing volume; more records will not fix weak targeting.

Iterate after the first output

Treat the first run as a calibration pass. Ask for a small batch, inspect false positives, then update the prompt with exclusion rules and preferred signals. For example: “The first batch included consultants; rerun with a rule to exclude agencies, independent consultants, and companies with fewer than 20 employees.” This creates better second-pass results than simply asking the agent to “try again.”

Add local workflow rules if your team needs consistency

Because the upstream skill has only SKILL.md, teams with repeatable lead research processes may want to add their own wrapper instructions: approved fields, naming conventions, deduplication rules, compliance notes, CRM import format, and review checkpoints. These additions make the icypeas-automation guide more durable for production use without changing the core requirement to discover live Rube tool schemas first.