Mihaela Cicvaric | September 4, 2026 | 17 min read
Mihaela Cicvaric | September 4, 2026 | 17 min read
Ready to launch?
Convert strategy into pipeline
Launch your first outbound campaign with verified leads, AI sequences, and multichannel outreach from one platform.
Sales Opportunity Detection with Claude Skills: A Practical 2026 Workflow Guide
What the Claap Sales Opportunity Detector does
The Claap Sales Opportunity Detector is a Claude Skill that scans your Claap call transcripts for buying signals and returns a ranked list of sales opportunities you might have missed. It’s built by lemlist, open-source, and free to download from the l3mpire/claude-skills GitHub repository.
A Claude Skill is a SKILL.md file, a packaged set of markdown instructions Claude loads automatically when your prompt matches its trigger description. You don’t install an app or learn new software. You ask Claude to find opportunities in your calls, and the skill takes it from there.
Under the hood, it connects to your Claap workspace through Claap MCP (Model Context Protocol), pulls transcripts in batch, and scores each one against a rubric of buying signals. The result is a scored, prioritized list of leads hiding in calls your team already had.
Who gets the most out of it:
- Sales teams using Claap wanting to surface upsell, expansion, or new pipeline buried in old recordings
- RevOps and sales leaders who’d rather run one batch scan than review calls one by one
- GTM engineers wiring Claap transcript analysis into broader AI agent workflows
How the signal detection grid scores each call
Every transcript gets checked against a three-tier rubric, and each tier carries a different point weight. Only statements the prospect makes count. Anything the rep says gets ignored, since the point is to measure real buyer intent, not the pitch.
Explicit product interest and tool dissatisfaction
Tier 1 signals carry the highest weight, 3 points each. These cover direct interest (“how does that work exactly?”), complaints about a current note-taker or recording tool, requests for automatic CRM filling, and phrases like “we’re evaluating solutions.” Calling a current tool unreliable or overpriced fires this tier too.
Process pain and coaching gaps
Tier 2 signals carry 2 points each: wanting to add AI to the sales process, trouble finding or sharing past recordings, coaching gaps, weak follow-up habits. “Managers don’t have time to listen to calls” is a textbook Tier 2 signal.
Competitor mentions and growth indicators
Tier 3 signals carry 1 point each, the lightest weight in the grid. Mentioning a competitor (Gong, Chorus, Modjo, Loom), general curiosity about sales tools, or hiring and expansion talk all land here. Weak individually, but a few together add up fast.
How the score maps to opportunity priority
Once a call finishes scoring, the skill totals the points and assigns a priority label.
Score range | Priority label | Recommended action |
|---|---|---|
6+ | Hot opportunity | Reach out within 48 hours |
3–5 | Warm opportunity | Nurture and monitor the next call |
1–2 | Weak signal | Add to a watch list |
0 | No signal | Excluded from the report |
Calls scoring zero get dropped entirely, so what lands in your report is already filtered to calls worth your attention.
How to install and run the skill
Two things need to exist first: a Claap account with recorded calls, and a working Claap MCP connection to Claude. Once both are in place, setup takes a few minutes.
1. Connect Claap MCP to Claude
Claap MCP is a Model Context Protocol server that gives Claude access to your Claap data: recordings, transcripts, deals, contacts, companies. Open Claude’s connector settings, add the Claap MCP endpoint, and sign in with OAuth. Claude usually registers the five Claap MCP tools in under two minutes, and without this step, the skill has nothing to scan.
2. Download the SKILL.md file
Grab the raw file from GitHub at l3mpire/claude-skills/claap-sales-opportunity-detector/SKILL.md. A Claude Code skill lives under ~/.claude/skills// and tells Claude how to perform a specific task, in this case scanning Claap calls. Or skip the manual placement and run the one-command install:
npx skills add l3mpire/claude-skills/claap-sales-opportunity-detector.3. Run your first scan
The skill opens with a short clarification phase, asking a few setup questions. Answer them, or skip and let it fall back to defaults.
Setting | Default if you skip |
|---|---|
Date range | Last 30 days |
Workspaces | All available |
Max calls to scan | 50 (most recent) |
Output format | CSV + inline table |
Filter by rep or team | None (all) |
After that, the skill fetches transcripts through MCP, scores them, and returns results, all inside a single Claude conversation.
Prompts that trigger the Claap Sales Opportunity Detector
There’s no special command to memorize. The skill activates whenever your prompt matches its trigger phrases, in French or English, and Claude loads it on its own.
A few prompts that work:
- “Find sales opportunities in my Claap calls from the last two weeks”
- “Who mentioned switching away from their note-taker in recent recordings?”
- “Scan my Claap transcripts for buying signals and export a CSV”
- “Analyse mes transcripts pour trouver des leads”
Adding account names, date ranges, or specific signal types to your prompt narrows the scan. More context tends to sharpen the output.
Reading the ranked opportunity table and CSV output
A finished scan returns two things: an inline table sorted by score, highest first, and a downloadable CSV. Here’s what each field means:
- nom_entreprise: company name, from recording metadata or Claap’s company search
- nom_prospect: the main non-rep participant on the call
- score: total signal points for that call
- niveau: priority level (hot, warm, or weak signal)
- signaux_detectes: the specific signal IDs that fired, like T1-B or T2-A
- intent_detecte: a one-line summary of what got detected
- citation_precise: the exact prospect quote behind the score, never paraphrased
- date_call and lien_recording: call date and a direct link back to the recording
Below the table, the skill prints a short summary: calls analyzed, calls with signals, breakdown by priority tier. The CSV is UTF-8 encoded and named with the scan date, so it drops straight into a spreadsheet or another tool with no reformatting.
Turning detected opportunities into outbound campaigns
A ranked list only matters once someone acts on it. The next move is building outbound around the exact signal the skill flagged.
Adding scored leads to a multichannel sequence
Export the CSV and upload it to lemlist, or connect lemlist MCP and push leads directly into a campaign without the export step. The signal data (intent summary, verbatim quote, priority level) becomes custom variables inside that campaign. From there, sequences can run across email, LinkedIn, calls, and SMS in one workflow.
Personalizing outreach with call context
This is where the workflow earns its keep. Instead of a generic opener, your first message can quote what the prospect actually said on a recorded call. lemlist’s enrichment agents layer on additional context from LinkedIn, company sites, and CRM records, so the message combines what they said with who they are.
💡 Tip: Running lemlist alongside Claap through MCP lets you go from signal detection to a live campaign inside a single Claude conversation. No CSV export required.
How this skill compares to Claap’s native deal scoring
Claap already analyzes every customer interaction to surface buying signals, objections, and deal risk on its own. So why add a separate skill? Because the two tools solve different problems.
Claap Sales Opportunity Detector (Claude Skill) | Claap native revenue intelligence | |
|---|---|---|
Scope | Batch scan across many calls at once | Per-deal, real-time monitoring |
Output | Ranked CSV, campaign-ready lead list | Dashboard view inside Claap |
Best for | Finding hidden pipeline in past recordings | Tracking deal health, coaching reps |
Customization | Edit the SKILL.md to change signal criteria | Configure scoring in Claap settings |
Action layer | Feeds into outbound tools like lemlist | Syncs to HubSpot or Salesforce fields |
They’re not fighting for the same job. Use the skill to sweep back through your call library for pipeline you missed. Use Claap’s native scoring to keep tabs on deals already moving.
Customizing the signal criteria for your product and ICP
The default SKILL.md is tuned for Claap’s own sales motion, which means it’s probably not a perfect fit for yours. That’s fine, the file is meant to be edited.
Editing signal keywords in the SKILL.md file
Open the SKILL.md, find the three signal tier tables, and swap in your own keywords. Selling project management software instead of a call recorder? Replace “Gong/Fireflies/Otter” with your category’s competitors, and swap CRM-filling pain points for whatever your product actually solves. The scoring logic stays the same regardless of the words plugged in.
Adding or removing signal tiers
A Tier 0 for deal-closing signals, budget confirmation or timeline commitment, is an easy addition. Tier 3 can go if competitor mentions create too much noise. Point weights are adjustable too: a team selling on urgency might weight growth signals heavier, while one running expansion plays might push tool dissatisfaction to the top.
Start detecting sales opportunities in your Claap calls today
The full loop: record calls in Claap, install the skill, run a scan, act on what it finds. Feed the output into lemlist and outreach can go out across channels, built around real call context instead of a guess.
Start a 14-day free trial to see how lemlist turns detected signals into live campaigns.
Frequently asked questions
Does the Claap Sales Opportunity Detector work with recordings from other platforms like Gong or Modjo?
No. The skill only fetches transcripts through the Claap MCP, so it’s limited to recordings already in your Claap workspace. lemlist’s Claude Skills library has separate skills for transcripts pasted in from other sources.
Can multiple team members run the skill on the same Claap workspace?
Yes, as long as each person has their own Claude environment connected to Claap MCP. The MCP is a permissioned window into data you can already see, not a shortcut around access controls, so each scan only covers recordings that person can view.
How often is it worth running opportunity detection scans?
Weekly or biweekly works for most teams. The skill defaults to a 30-day window, so a weekly cadence catches new recordings while overlapping with calls already scanned. High call-volume teams might run it more often with a shorter window.
Does the skill detect buying signals in languages other than English?
Yes. It works in whatever language the call was recorded in, and the default SKILL.md already includes French and English trigger phrases. Keywords for other languages can be added to the signal tables if your team runs calls in more than those two.
What the Claap Sales Opportunity Detector does
The Claap Sales Opportunity Detector scans Claap transcripts to detect buying signals and surface ranked sales opportunities. (Claap Sales Opportunity Detector - lemlist) It’s a Claude Skill, which is a SKILL.md file (a packaged set of markdown instructions) that Claude auto-loads when your prompt matches its trigger description. It analyzes Claap recordings using a 3-tier signal detection grid, scoring each call and producing a ranked opportunity table plus a CSV export (Claap Sales Opportunity Detector - lemlist) with verbatim citations.
The skill is open-source, hosted in lemlist’s GitHub repository (l3mpire/claude-skills), and free to download. It connects to your Claap workspace through the Claap MCP (Model Context Protocol), pulls call transcripts in batch, and runs them against a structured rubric of buying signals. The output is a scored, prioritized list of opportunities hiding in calls your team has already had.
Who gets the most from it:
- Sales teams using Claap who want to uncover upsell, expansion, or new pipeline leads buried in existing recordings
- RevOps and sales leaders who want a repeatable, batch-level scan instead of reviewing calls one by one
- GTM engineers building AI agent workflows that connect Claap transcript analysis to outbound action
How the signal detection grid scores each call
The skill evaluates every transcript against a structured rubric with three tiers. Each tier carries a different point weight, and the skill only flags statements made by the prospect, never the rep. Here’s how each tier works.
Explicit product interest and tool dissatisfaction
These are Tier 1 signals, and they carry the highest weight at 3 points each. The skill looks for direct expressions of interest (“how does that work exactly?”), stated dissatisfaction with a current note-taker or recording tool, requests for automatic CRM filling, and active tool evaluation language like “we’re evaluating solutions.” If a prospect says their current tool is unreliable or too expensive, that fires a Tier 1 signal.
Process pain and coaching gaps
Tier 2 signals carry 2 points each. These capture prospects who want to add AI to their sales process, struggle with sharing or finding past call recordings internally, have rep coaching gaps, or deal with poor follow-up workflows. A prospect saying “managers don’t have time to listen to calls” would trigger this tier.
Competitor mentions and growth indicators
Tier 3 signals carry 1 point each. Any mention of a competing tool (Gong, Chorus, Modjo, Loom), general curiosity about sales tools, or hiring and market-expansion activity flags here. These are weaker individually but compound quickly across multiple mentions.
How the score maps to opportunity priority
After scanning, the skill sums points per call and assigns a priority label:
Score range | Priority label | Recommended action |
|---|---|---|
6+ | Hot opportunity | Reach out within 48 hours |
3–5 | Warm opportunity | Nurture and monitor the next call |
1–2 | Weak signal | Add to a watch list |
0 | No signal | Excluded from the report |
Only calls scoring 1 or above appear in the output. Zero-signal calls are dropped entirely.
How to install and run the skill
You’ll need two things already in place: a Claap account with recorded calls, and a working Claap MCP connection to Claude. Once those are set, the setup takes a few minutes.
1. Connect Claap MCP to Claude
Claap MCP is a Model Context Protocol server that connects your Claap call data (recordings, transcripts, deals, contacts, and companies) to Claude. (Claap MCP: the most advanced sales MCP for Claude, ChatGPT, Cursor) Open Claude’s connector (MCP) settings, add the Claap MCP endpoint, and sign in with OAuth. Claude detects and registers the five Claap MCP tools automatically, usually in under two minutes. (Claap MCP: the most advanced sales MCP for Claude, ChatGPT, Cursor) The skill won’t work without this connection.
2. Download the SKILL.md file
You can grab the raw file from GitHub at
l3mpire/claude-skills/claap-sales-opportunity-detector/SKILL.md. A Claude Code skill is a SKILL.md file that lives under ~/.claude/skills/<name>/ and tells Claude how to perform a specific task. (Sales Claap — Claude Code Skill | ClaudSkills) Place the file in the correct directory, or use the one-command install: npx skills add l3mpire/claude-skills/claap-sales-opportunity-detector.3. Run your first scan
The skill opens with a clarification phase, asking a few setup questions before it begins. You can answer or skip to accept the defaults:
Setting | Default if you skip |
|---|---|
Date range | Last 30 days |
Workspaces | All available |
Max calls to scan | 50 (most recent) |
Output format | CSV + inline table |
Filter by rep or team | None (all) |
After clarification, the skill fetches transcripts via MCP, scores each one, and returns results. The whole scan runs inside a single Claude conversation.
Prompts that trigger the Claap Sales Opportunity Detector
You don’t type a specific command. The skill activates when your prompt matches certain trigger phrases, and Claude loads it automatically. The skill triggers when the user wants to find leads hidden in existing customer or prospect calls (Claude Code Skills Directory - 100+ AI Agent Skills | ColdIQ), even in French or English.
Some ready-to-use prompts:
- “Find sales opportunities in my Claap calls from the last two weeks”
- “Who mentioned switching away from their note-taker in recent recordings?”
- “Scan my Claap transcripts for buying signals and export a CSV”
- “Analyse mes transcripts pour trouver des leads”
You can scope the scan further by adding account names, date windows, or specific signal types to the prompt. The more context you give, the more targeted the output.
Reading the ranked opportunity table and CSV output
The skill returns two things: an inline summary table sorted by score (highest first) and a downloadable CSV. Here’s what each field means:
- nom_entreprise: company name, pulled from recording metadata or Claap company search
- nom_prospect: the main non-rep participant on the call
- score: total signal points for that call
- niveau: priority level (hot, warm, or weak signal)
- signaux_detectes: list of signal IDs that fired (e.g., T1-B, T2-A)
- intent_detecte: one-line natural-language summary of what was detected
- citation_precise: the exact prospect quote justifying the score, never paraphrased
- date_call and lien_recording: call date and direct link back to the Claap recording
After the table, the skill prints a scan summary showing calls analyzed, calls with signals, and counts per priority tier. The CSV is a ready-to-use file with 10 fields sorted by score (Claap Sales Opportunity Detector - lemlist), encoded in UTF-8 and named with the scan date.
Turning detected opportunities into outbound campaigns
A ranked opportunity list is only useful if someone acts on it. The natural next step is to take those scored leads and build outbound around the exact signal that was detected.
Adding scored leads to a multichannel sequence
Export the CSV from the skill and upload it to an outbound platform like lemlist, or use lemlist MCP to push leads directly into a campaign. The signal data (intent summary, verbatim quote, priority level) becomes custom variable fields. From there, you can launch automated sequences across email, LinkedIn, calls, and SMS from one workflow, with every message tied to real context from the call.
Personalizing outreach with call context
The verbatim quotes and intent summaries are what make this workflow different from generic outbound. Instead of a templated opener, the first message can reference what the prospect actually said on a recorded call. lemlist’s data enrichment agents can also pull additional context from LinkedIn, websites, and CRM data to layer on top of the Claap signal. The loop is: record → detect → score → personalize → send.
💡 Tip: If you’re running lemlist alongside Claap, you can connect both through MCP and go from signal detection to live campaign in a single Claude conversation, without exporting a CSV at all.
How the skill compares to Claap’s native deal scoring
Claap automatically analyzes every customer interaction to surface buying signals, objections, and deal risks or opportunities. (Revenue Intelligence Platform for Sales Teams) So you might be wondering: why run a separate skill?
The two tools do different jobs:
Claap Sales Opportunity Detector (Claude Skill) | Claap native revenue intelligence | |
|---|---|---|
Scope | Batch scan across many calls at once | Per-deal, real-time monitoring |
Output | Ranked CSV + campaign-ready lead list | Dashboard view inside Claap UI |
Best for | Finding hidden pipeline in existing recordings | Tracking active deal health and coaching reps |
Customization | Edit the SKILL.md to change signal criteria | Configure scoring frameworks in Claap settings |
Action layer | Feeds into outbound tools like lemlist | Syncs to HubSpot or Salesforce CRM fields |
They’re complementary. Use the skill for retroactive batch analysis across your full call library. Use Claap’s built-in scoring for live deal management.
Customizing the signal criteria for your product and ICP
The default SKILL.md is tuned for Claap’s own sales team, but you can (and probably will want to) edit it to match your product and market.
Editing signal keywords in the SKILL.md file
Open the SKILL.md, find the three signal tier tables, and replace the example keywords with your own. If you sell a project management tool instead of a call recorder, swap “Gong/Fireflies/Otter” for the competitor names in your space. Replace CRM-filling signals with the pain points your product solves. The detection grid works the same way regardless of what keywords it’s scanning for.
Adding or removing signal tiers
You can add a Tier 0 for deal-closing signals (budget confirmation, timeline commitment) or remove Tier 3 entirely if weak signals generate too much noise. Adjust point weights to match your sales motion. A team that closes on urgency might weight growth signals higher, while a team running expansion plays might weight tool dissatisfaction at the top.
Start detecting sales opportunities in your Claap calls today
The full workflow is: record calls in Claap → install the skill → run a scan → act on the scored leads. If you want to go further, connect the output to lemlist and launch multichannel outreach personalized with real call context.
Start a 14-day free trial or book a demo to see how lemlist turns detected signals into live campaigns.
Frequently asked questions
Does the Claap Sales Opportunity Detector work with recordings from other platforms like Gong or Modjo?
No. The skill always fetches transcripts via the Claap MCP (Claude Code Skills Directory - 100+ AI Agent Skills | ColdIQ), so it only accesses recordings stored in your Claap workspace. lemlist’s Claude Skills library includes separate skills for analyzing transcripts pasted from other sources.
Can multiple team members run the skill on the same Claap workspace?
Yes, as long as each person has their own Claude environment with Claap MCP connected. The MCP is a permissioned window into the data you can already see, not a back door. (Claap MCP: the most advanced sales MCP for Claude, ChatGPT, Cursor) Each user’s scan is scoped to the recordings they have access to in Claap.
How often is it worth running opportunity detection scans?
A weekly or biweekly cadence works well for most teams. The skill defaults to scanning the last 30 days, so running it weekly catches new recordings while overlapping with previously scanned calls. Teams with high call volume might run it more frequently with a shorter date window.
Does the skill detect buying signals in languages other than English?
Yes. The skill handles transcripts in whatever language the call was recorded in. The default SKILL.md includes both French and English trigger phrases and keywords. You can add keywords in other languages to the signal detection tables if your team runs calls in additional languages.
Content Marketing Manager @lemlist
Share this post