What are skills in Cleanlist?

Skills run AI work on each row of a lead list: the Research skill, the Qualification skill, the Custom skill and Employment check, plus any prompts your team saved. Run them from a list's Actions drawer, on the Skills tab. Agents can also run skills inside their plays.

What happened to the Research, Qualification and Custom agents?

They are now called skills: the Research skill, the Qualification skill and the Custom skill, on the Skills tab of a list's Actions drawer. They work the same way. The word Agents now means automations that run plays.

How much do skills cost?

Each skill's price per row comes from an estimate shown in the Actions drawer before you run, and that is what the run charges. Default prices are about 6.3 credits for a contact, 12.2 for both and 17.8 for a company with the Research skill, about 0.6 with the Qualification skill, and about 0.2 to 0.6 with the Custom skill depending on the prompt. Employment check costs up to 3.5 credits per person.

Which plan do I need to run skills?

Starter, Pro or Enterprise, including the Pro trial. The Free plan cannot run skills. AppSumo workspaces include the Research, Qualification and Custom skills but not Employment check.

Does the Research skill search the web?

Yes. For each row it can run up to 6 web searches, read full pages and check your connected CRM, and it must take at least two investigation steps before it answers. It runs for at most eight turns.

Which CRMs does the Qualification skill check?

Every CRM your workspace has connected among HubSpot, Salesforce and Lemlist, in parallel, with a 5 second timeout each. If no CRM is connected, the result says so instead of guessing.

Can I stop the Qualification skill from scoring leads I never want?

Yes. Add titles to the excluded personas and countries, states or cities to the excluded geography on the Agents page Profile tab. A matching lead is disqualified right away with the reason, without a model call, but it is still part of the run's quoted price.

What does Employment check do?

It checks whether each person still works at the company on the row and returns Yes, No or Unknown with a confidence, plus where leavers went, their new title, an updated email when found, and the status of the email on file. It runs on the Leads sheet, up to 1,500 people per run, and costs up to 3.5 credits per person.

How do I save a Custom skill so my team can reuse it?

In the Custom skill run panel, turn on Save to skills and give it a name, or create one on the Agents page with Skills, New skill, Per record. Only organization admins can save, edit or delete saved skills, and saved skills appear under Your skills in every list's Skills tab.

How do I use lead fields in a skill prompt?

Type / in the prompt editor and pick a field; it is inserted as a variable chip such as {{lead.first_name}}. In a Custom skill you can also write a field's label after a slash, such as /Company Name. Text after a slash that matches no field stays as written.

What is the Organization Profile?

A summary of your company that Cleanlist builds from your website and, optionally, your company LinkedIn page: what you sell, your ICP, personas to target and exclude, geography, value props, competitors and products. The Research and Qualification skills read it on every run. Building it is free.

Can a skill run automatically on new leads?

Skills on a list run when you start them. After adding leads, choose Run on all unfilled rows to process only the rows without a result. To process new leads automatically, use an agent on Pro or Enterprise with the Leads added to a list trigger and an Agents step that runs the skill.

Are there limits on skill runs?

Each workspace can start up to 20 skill runs per minute, and a prompt can be at most 8,000 tokens (the editor warns from 2,000). Employment check takes up to 1,500 people per run and one run per list at a time. Rebuilding the Organization Profile is limited to 6 times per hour.

👤

Who this is for: self-serve plans (Free, Starter, Pro, Enterprise). If you bought a Cleanlist AppSumo lifetime deal, your features, limits, and pricing are different. See the AppSumo documentation instead. Support agent: confirm whether the user is self-serve or AppSumo before giving plan-specific advice.

TL;DR: Open a lead list, click Actions, and pick the Skills tab. Research skill (live web research), Qualification skill (ICP score plus CRM check), Custom skill (your prompt) and Employment check (who still works where). The price per row is shown before you run; defaults are about 6.3 / 12.2 / 17.8 for Research (contact / both / company), 0.6 for Qualification, 0.2 to 0.6 for Custom, and up to 3.5 for Employment check. Starter and above, including the Pro trial. These used to be called agents.

Skills on lead lists

Skills do the reading and judging a rep would otherwise do by hand. You point one at a lead list, it works through the rows in the background, and it writes the results back as columns.

The Research, Qualification and Custom skills used to be called "agents". In Cleanlist today, Agents are automations that run plays, and the per-row tools are skills. Agents can run the same skills inside a play with the Agents step.

The skills

Skill	What it does	Works on	Price per row
Research skill	"Deep-research the company, contact, or both. Personalized to your ICP."	Leads and Companies sheets	From your estimate. Defaults about 6.3 (contact), 12.2 (both), 17.8 (company)
Qualification skill	"Score against your ICP + auto-check whether they are already in your CRM."	Leads and Companies sheets	From your estimate. Default about 0.6
Custom skill	"Write your own prompt. Cleanlist context injected automatically."	Leads and Companies sheets	From your prompt's estimate. About 0.2 to 0.6
Employment check	"Check whether each person still works at their company, and where they went."	Leads sheet only	Up to 3.5, billed after the check
Your skills	Custom prompts your team saved	Leads and Companies sheets	From the prompt's estimate

Who can use skills

Plan	Research, Qualification, Custom	Employment check
Free	No	No
Starter	Yes	Yes
Pro and Enterprise	Yes	Yes
Pro trial	Yes	Yes
AppSumo	Yes, see AppSumo	No

On plans without Agents (Starter and AppSumo), the Agents page shows two tabs: Catalog, with the built-in skills and Your custom agents (Add custom agent), and Profile.

How skill prices work

Skill prices are no longer flat. Each skill's price per row is estimated from the AI work it needs (and the web searches it expects), rounded up to the nearest 0.1 credit.

The Skills tab shows each skill's price for your workspace. Research shows a range across its scopes.
The run panel shows the exact price per scope ("Company = X credits/lead. Contact = Y credits/lead. Both = Z.") and the per-row price on the run button.
A run's total is rounded up once to a whole credit, reserved when the run starts, and charged as quoted.
A Custom skill's price depends on the length of its prompt.
Employment check is different: nothing is charged at the start, and each row is billed after the check for the lookups it made: email status 0.5 and reverse email lookup 1 whenever they run, profile lookup 1 only when the person is identified, and a new email 1 only when found. That is at most 3.5 credits per person.

Building or rebuilding the Organization Profile is always free.

Run a skill on a list

Open a lead list and click Actions (⌘A), or click the + column header (Add column).
Pick the Skills tab and choose a skill.
For the Research skill, choose the Research scope: Company, Contact or Both. On the Companies sheet the scope is fixed to company.
Review the prompt. Built-in skills show Prompt (edit per-run): changes apply to this run only. The Custom skill shows Your prompt. Generate drafts a prompt from a plain-English description, and a token counter warns when the prompt gets long.
Optional settings:
- Save to skills (Custom skill, admins): saves the prompt with a name so anyone can reuse it.
- Invalid email counts as left (Employment check): "Only a work email on the company's own domain. Marks the person as left (medium confidence), even if their profile says otherwise." Off by default.
- Only run if conditions, to leave out rows that don't match.
Open the Run menu on the button (it shows the price per row) and pick Run on N rows, Run on all N unfilled rows, Re-run all N rows, or Save settings only (keep the settings without running now).

Results stream back row by row. Large lists run in the background, so you can close the tab. See Actions drawer for the other tabs.

Read the results

Open any skill cell to see:

The written analysis
Structured insights, where you can promote any single field to its own column (Promote to column)
Agent trace (Research skill), showing each search and page read. Trace excerpts are redacted of personal data and shortened.

Research skill

The Research skill investigates each row on the live web, then writes back a structured record and a readable analysis.

How it works. For each row it can run up to 6 web searches, read full pages, and check your connected CRM. It must complete at least two investigation steps before it is allowed to answer, and it runs for at most eight turns. It reads your Organization Profile so its output is tuned to what you sell.

Company scope returns: company name, one-liner, industry, employee-count estimate, founded year, HQ location, products, positioning, recent news (up to 5), hiring signals, tech stack, funding rounds (up to 8) and competitors (up to 10).

Contact scope returns: full name, current role, tenure in months, prior companies (up to 8), areas of responsibility, public content (up to 8) and talking points.

Both returns the two together.

Qualification skill

The Qualification skill answers two questions in one pass: is this lead worth working, and do you already have them in your CRM. It runs in three stages.

Exclusion check, no model call. If the lead's job title matches an excluded persona, or their country, state or city matches an excluded location in your Organization Profile, the lead is disqualified immediately with fit_score: 0 and the reason. Exclusions beat inclusions. The row is still part of the run's quoted price.
CRM check. Cleanlist looks the person up in every CRM your workspace has connected among HubSpot, Salesforce and Lemlist, in parallel, with a 5 second timeout each, and takes the first match.
ICP scoring. The model scores the lead against your Organization Profile. The real CRM result is written over whatever the model says, so the CRM field is never a guess.

Field	Values
`fit_score`	0 to 100
`verdict`	`qualified`, `disqualified`, `needs_review`
`verdict_reasons`	1 to 4 short reasons
`crm_status`	A match (CRM, record link, record ID and owner), a confirmed absence, or a lookup problem (no CRM connected, or timeout)
`disqualification_flags`	Any of `competitor`, `partner`, `current_customer`, `sub_icp_size`, `geo_mismatch`, `wrong_persona`, `no_business_signal`
`analysis`	Written rationale

The Qualification skill does not search the web.

Custom skill

Write a prompt once and run it on any list. The Custom skill sees the row's lead, contact and company data ("Cleanlist context injected automatically").

Output. Without declared outputs, the skill returns one value (text, a list, yes/no, a number, or key-value pairs) chosen from your prompt. A saved per-record skill can declare Outputs: up to 12 named fields typed Text, Number, Yes/No, Choice, List or Date, which later steps and columns can use.

Saving. Turn on Save to skills in the run panel, or go to Agents → Skills → New skill → Per record (on plans without Agents: Agents → Catalog → Add custom agent). The dialog may still be titled "New custom agent". Give it a Name, an optional Description, and the Prompt. Saving, editing and deleting are admin only ("Only org admins can create custom agents."), and names must be unique. Saved skills appear under Your skills in every list's Skills tab. Deleting one keeps every result it already produced.

What the Custom skill does not do. It does not search the web, read pages, query your CRM or read your Organization Profile. It sees only the data already on the row. For live research, use the Research skill.

Employment check

Employment check finds out whether each person still works at the company on the row.

Result: Still at Company? Yes, No or Unknown, with a Confidence.
Extra columns: Current Employer, Current Title, Left On, Updated Email, Email Status, LinkedIn URL, Evidence Source, Evidence URL, Last Verified Date and Notes.
Runs on the Leads sheet only (people, not companies), up to 1,500 people per run, one run per list at a time.
Costs up to 3.5 credits per person, billed per row after the check for the lookups it made (see How skill prices work).
Invalid email counts as left (run panel switch, off by default) treats an invalid work email on the company's own domain as a departure, with medium confidence.
Not included in AppSumo-only workspaces.

To recheck a whole CRM every quarter automatically, use an agent instead. See the still-employed recipe in Agent recipes.

Use row data in a prompt

Type / in the prompt editor and pick a field. It appears as a chip and is sent as a variable such as {{lead.first_name}}. Fields come in groups:

Group	Examples
Lead	First name, Last name, Full name, Email, Job title, LinkedIn URL, Location
Prospect	LinkedIn URL, LinkedIn headline, LinkedIn about, Prospect location
Company	Company name, Website, Company LinkedIn, Industry, Employee count, HQ location, Company description
Your company	Your company name, Your one-liner, Your ICP and other Organization Profile fields

Research and Qualification fill every group, including your Organization Profile.
Custom skill fills the Lead, Prospect and Company variables. It also accepts a field's display label after a slash, matched without regard to capitals: /First Name, /Job Title, /Email, /Company Name, /Company Domain, /LinkedIn URL, /Industry, /Company Size, /City, /State, /Country, /Full Location, /Created Date, plus the name of any column on your list.
A missing value reads as empty. Text after a slash that matches no field (a URL path, "and/or") stays as written, and so do {{ }} tokens from other tools, such as sequencer merge tags.

Test a new prompt on a handful of rows and read the output before running it on a whole list.

Organization Profile

The Research and Qualification skills are grounded in an Organization Profile, a structured summary of your company. Until someone in your workspace completes it, the Agents page shows a one-time setup instead.

Set it up: on the Agents page, enter your Company website and, optionally, your Company LinkedIn, then click Build my profile. Cleanlist reads your site and researches your company in the background ("This takes a few minutes. Keep working"; you can close the page). Review the draft under Looks right? (Identity, Who you sell to, Positioning) and click Looks good, save my profile.

What it holds (on the Profile tab, "Your agents profile"): Company name, Industry, One-liner, Ideal Customer Profile, personas to target and personas to exclude, geography (Countries, States / Provinces, Cities, each with include and exclude lists), Value props, Competitors (up to 10), Products, Website and LinkedIn.

Free, every time, including rebuilds. Rebuilds are limited to 6 per hour per workspace.
One person unlocks it for everyone. First-time setup is open to any member.
Re-run research and Edit profile are admin only.
The excluded personas and excluded locations are the only part of the profile that can disqualify a lead before any model runs, which makes them the cheapest filter you have.
The Profile tab also has a switch, Use my ICP to fill in unclear requests, that lets Clu use the profile when a request is vague.

Skills library

On plans with Agents, Agents → Skills lists every skill:

Kind: Per record (runs once per row or record, like the skills on this page) or Procedure (written instructions an agent loads when a task needs it: Name, Description "When the agent should load it", and Instructions).
Used by: the agents and plays that use the skill.
Row menu: Edit prompt (built-in Research and Qualification, to change your workspace's default prompt; admins), Edit, Use in list (pick a list to run it on) and Delete.
New skill → Procedure is open to members; New skill → Per record is admin only.

On plans without Agents, admins change the default prompts from Agents → Catalog → Edit prompt.

Skills in agents and sequences

Plays: the Agents step runs these skills on each record, plus Write and Agent instructions. See Play steps reference.
Sequences: a call or manual task step can run a skill on each lead who reaches it, priced like the skill. See Sequences.
Clu can run skills on a list for you; it shows the price on a confirmation card first. See Clu.

Limits

Limit	Value
Skill runs started per workspace	20 per minute
Prompt length	8,000 tokens (warning from 2,000)
Employment check	1,500 people per run, one run per list at a time
Organization Profile rebuilds	6 per hour

Actions drawer: Import, Enrich, Skills and Export
Agents: automations that run skills inside plays
Agent recipes: includes recipes for skills on a list
Credits and pricing
Integrations: the CRMs the Qualification skill checks
Tutorials: videos on the Research and Qualification skills

Play Steps Reference Knowledge