Remote OpenClaw Blog
OpenClaw Lead Scorer: Score and Prioritize Leads Automatically
10 min read ·
Remote OpenClaw Blog
10 min read ·
The OpenClaw Lead Scorer is a free skill from the Remote OpenClaw marketplace that takes a CSV file of leads, compares each one against your defined Ideal Customer Profile (ICP), and returns a scored, categorized list with transparent reasoning for every rating. It runs entirely on your OpenClaw instance using the LLM model you already have configured -- no additional API keys, no third-party enrichment services, and no subscription fees.
Lead scoring is one of the most time-consuming steps in any sales workflow. Manually reviewing a list of 200 leads from a conference, a scraping tool, or a marketing campaign can take an entire afternoon. Lead Scorer processes that same list in minutes and surfaces the 15 to 30 leads that are worth contacting first.
The skill was designed for founders, solo sales operators, and small teams who generate leads faster than they can qualify them. It does not replace human judgment -- it accelerates it by doing the initial sort and providing written reasoning that you can scan, challenge, and override (OpenClaw GitHub repository).
Lead Scorer uses a weighted comparison model. You define your ICP -- the characteristics of your ideal buyer -- and the skill evaluates each lead against those criteria. Criteria that match increase the score; criteria that conflict decrease it; missing data is noted but does not penalize the lead.
The scoring happens in three phases:
This approach is fundamentally different from rules-based lead scoring tools that assign fixed points per field match. The LLM-powered evaluation can interpret nuance -- for example, understanding that a "VP of Revenue Operations" at a 50-person SaaS company is a strong match for a sales automation ICP, even though the title does not contain the word "sales."
Your ICP definition is a plain-text description of your ideal buyer. Lead Scorer reads this definition before processing each batch and uses it as the reference standard for all scoring decisions.
A strong ICP definition includes:
The ICP definition is stored in a text file on your OpenClaw instance. You can update it at any time, and the next scoring run uses the updated definition. Most operators refine their ICP after the first two or three batches as they see which scoring decisions align with their judgment and which need adjustment.
Lead Scorer works with standard CSV files exported from LinkedIn Sales Navigator, Apollo, Hunter.io, conference badge scanners, Google Sheets, Excel, or any tool that exports to CSV. It does not require a specific column order or naming convention.
The column auto-detection system recognizes common field names and their variations:
If your CSV uses non-standard column names, Lead Scorer asks you to confirm the mapping before processing. This happens once per unique CSV format -- subsequent files with the same headers are processed automatically.
Not every column needs to be present. Lead Scorer scores based on whatever data is available and explicitly notes which ICP criteria could not be evaluated due to missing columns. A CSV with only company name, title, and industry still produces useful scores -- they are just less precise than a CSV with complete data.
Lead Scorer produces a scored CSV file with all original columns preserved plus three new columns appended: lead_score (0-100), lead_category (Hot/Warm/Cold), and lead_reasoning (plain-English explanation).
The output file is sorted by score in descending order, so the highest-rated leads appear at the top. This makes it immediately usable -- open the file, start with row 1, and work down.
The original data is never modified. Lead Scorer reads your input CSV, processes it, and writes a new output file. Your source file remains untouched in case you want to re-run the scoring with an updated ICP or compare results across different ICP definitions.
Every scored lead is automatically categorized into one of three tiers based on the numerical score:
The category thresholds are configurable. If your ICP is narrow and you want a more selective Hot tier, you can raise the threshold to 85 or 90. If your ICP is broad and you want more leads in the contactable range, you can lower the Warm floor to 40.
The category system exists because a numerical score alone does not drive action. Sales operators need a fast answer to "should I contact this lead today?" -- and Hot/Warm/Cold provides that answer without requiring them to interpret a number.
Lead Scorer is designed for batch operations. It processes CSV files containing anywhere from 10 to 500+ leads in a single run. Processing speed depends on your LLM model and API rate limits, but a typical batch of 200 leads completes in 3 to 8 minutes.
The skill processes leads in parallel chunks to maximize throughput while staying within API rate limits. It handles interruptions gracefully -- if the process is stopped mid-batch, the partially-scored results are saved and the skill can resume from where it left off.
For operators who receive leads continuously (from web forms, scrapers, or marketing campaigns), Lead Scorer can be triggered automatically when a new CSV file appears in a designated directory. Combined with OpenClaw's scheduled tasks, this creates a hands-off scoring pipeline where new leads are scored within minutes of arrival.
LLM cost per batch is minimal. A batch of 100 leads typically costs under $0.15 in API tokens using Claude or GPT-4, making this significantly cheaper than third-party lead enrichment services that charge $0.10 to $0.50 per lead.
The most common complaint about lead scoring tools is that the scores feel arbitrary. A lead gets a 72, but you have no idea why -- and no way to challenge or calibrate the model. Lead Scorer addresses this by generating a written explanation for every score.
The reasoning column contains 2 to 4 sentences explaining the key factors behind the score. For example:
"Scored 84 (Hot). Strong industry match: B2B SaaS aligns directly with ICP. Company size of 45 employees falls within the 10-200 target range. VP of Sales title matches buyer persona. Location (Austin, TX) is within the US target geography. No disqualifying signals detected."
And for a lower-scored lead:
"Scored 31 (Cold). Industry mismatch: government consulting does not align with B2B SaaS ICP. Company size of 3,200 employees exceeds the 200-employee maximum. Title (Administrative Coordinator) does not match decision-maker persona. Geographic fit confirmed (US-based)."
This transparency serves two purposes. First, it lets you verify individual scores and catch errors -- if the reasoning cites incorrect data, you know the issue is in the source CSV, not the scoring logic. Second, it helps you refine your ICP definition. If you consistently disagree with how a particular criterion is weighted, you can adjust the ICP description to better reflect your actual preferences.
Lead Scorer installs like any other OpenClaw skill. Download the SKILL.md file from the Remote OpenClaw marketplace, place it in your OpenClaw skills directory, and restart the agent. The skill registers itself and becomes available immediately.
Before your first scoring run, you need to create your ICP definition file. The skill's documentation includes a template with prompts for each section (industry, company size, buyer role, geography, technology signals, and disqualifiers). Most operators complete this in 10 to 15 minutes.
To run a scoring batch, send your OpenClaw agent a message like: "Score the leads in leads-april.csv against my ICP." The agent locates the file, confirms the column mapping if it is a new format, processes the batch, and returns the scored output file.
The skill file is a single SKILL.md document. No compiled code, no dependencies, no external services. You can read the entire file before installing to verify what it does and what data it accesses.
Lead Scorer handles one step in the sales pipeline: initial qualification. For operators who want their OpenClaw agent to manage the entire pipeline -- from lead research through outreach sequences to CRM updates and follow-up scheduling -- the Scout persona extends this capability across the full workflow.
Scout includes Lead Scorer as a built-in capability, plus automated lead research (enriching scored leads with LinkedIn data, company news, and technology stack information), personalized outreach sequence generation, CRM synchronization (pushing scored and enriched leads to Salesforce, HubSpot, or Pipedrive), and follow-up scheduling based on engagement signals.
If lead scoring is the bottleneck in your pipeline and you want to automate the steps before and after it, Scout provides the end-to-end framework.
Related guides:
No. Lead Scorer uses your OpenClaw instance's existing LLM connection to analyze leads. It does not require separate API keys, third-party enrichment services, or paid subscriptions. The skill itself is free, and the LLM cost per batch of 100 leads is typically under $0.15.
Lead Scorer uses automatic column detection and works with most common CSV formats. It looks for columns containing company name, contact name, email, title/role, industry, company size, location, and website. It does not require all columns to be present -- it scores based on whatever data is available and notes which ICP criteria could not be evaluated due to missing data.
Lead Scorer compares each lead's available data against your defined Ideal Customer Profile (ICP). It evaluates criteria like industry match, company size fit, role seniority, geographic alignment, and technology stack overlap. Each criterion contributes to a weighted score from 0 to 100. The reasoning for each score is written in plain English so you can verify the logic and adjust your ICP definition if the scoring does not match your judgment.