lead-scoring
Score a lead AFTER contact — start from the pre-contact profile score and add what only a conversation reveals: engagement, confirmed deal size, the real decision path, timeline. Use when asked to rank leads you have already spoken to, prioritise a reply queue, decide who gets a call first, or recalibrate scoring bands against a real pipeline. Triggers include "score these leads", "prioritise the pipeline", "who do we call first", "rank the replies". For scoring a cold base nobody has spoken to, use `prospect-scoring`.
How do I install this agent skill?
npx skills add https://github.com/victor-shulga/outbound-engine-skills --skill lead-scoringIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill defines a logical framework for business lead scoring based on engagement and deal data. It includes a glossary with Ukrainian translations for localization. No malicious code, data exfiltration patterns, or obfuscation attempts were detected.
- Socketpass
No alerts
- Snykwarn
Risk: MEDIUM · 1 issue
What does this agent skill do?
Lead Scoring
Scoring splits in two, because the data splits in two.
- Pre-contact — what you can read: profile fit, services fit. That is
prospect-scoring. - Post-contact — what only a conversation establishes: whether they engaged, what the deal is actually worth, who really decides, when they would move.
This skill is the second half, and it starts from the first. A lead score that re-derives fit from scratch throws away the work already done and produces two different fit numbers for the same account.
Required tools
Install these before running the skill. It still runs without the ones marked Required, but the result gets noticeably worse: post-contact blocks score 0 because engagement, named budgets and timelines are not where the skill can read them.
| Tool | Used for | Required? | Without it |
|---|---|---|---|
| CRM (Twenty, HubSpot) | prospect profile carried over, deal values, decision path, stage, opt-outs | Required | you paste every lead by hand; calibration on a clean cohort is not possible |
| Outreach platform (HeyReach, Instantly) | replies, threads and engagement per lead | Recommended | engagement is scored from what you remember |
| Call recorder transcripts | budget, timeline and who signs, as the lead said it | Recommended | deal and timeline blocks score 0 unless you type them in |
| Google Sheets | the team's live scoring sheet and its bands | Recommended | bands stay the rubric defaults, not calibrated ones |
The structure
lead score = prospect profile (carried over, capped) + post-contact points (0 until contact)
A workable default — tune the weights, keep the shape:
| Block | Max | Available |
|---|---|---|
Prospect profile (from prospect-scoring, normalised) | 30 | before contact |
| Engagement — replied, opened a thread, asked something real, attended | 15 | after contact |
| Confirmed deal potential — a number they said, not one you inferred | 20 | after contact |
| Decision path — you know who signs and how they buy | 15 | after contact |
| Timeline — a date or a window they named | 10 | after contact |
| Access — you are talking to the person who decides, or to a champion who will carry it | 10 | after contact |
Each block on levels, so the score reads the same whoever fills it in:
| Block | Full | Partial | 0 |
|---|---|---|---|
| Prospect profile (30) | the prospect-scoring total, carried over | — | — |
| Engagement (15) | 15 · came to a call | 10 · asked something real · 5 · "send me info" | no reply |
| Deal potential (20) | 20 · named a budget at or above your minimum deal | 10 · named one below it, or only a range | no number named |
| Decision path (15) | 15 · you know who signs and how they buy | 8 · you know who signs, not the process | unknown |
| Timeline (10) | 10 · a date within 3 months | 5 · "this year" or later | no timeline |
| Access (10) | 10 · talking to the decision-maker | 5 · talking to a champion who will carry it | whoever happened to reply |
Score → action, as starting bands (recalibrate, see below): 60+ a call · 40–59 one qualifying question aimed at a block that scored 0 · below 40 nurture track · stop (confirmed non-target or opt-out) out of all campaigns whatever the total.
Hard ceiling rule. A lead with zero post-contact points cannot outrank one that has them, however good its profile. Cap uncontacted rows below the top band. Otherwise the queue fills with beautiful strangers and the people who actually replied sit behind them.
Bands come from your pipeline, not from the rubric
A rubric's thresholds are a guess until a real pipeline calibrates them. If the team's live sheet treats 60+ as worth a call and the rubric says 75, the sheet wins — it is the one that has been compared against outcomes. Write both numbers down and reconcile them deliberately; do not let a theoretical threshold quietly override observed behaviour.
Recalibrate after every meaningful cohort, and recalibrate on accounts, not rows — a company with three contacts otherwise counts three times and drags the whole model.
Only calibrate on a clean cohort. Accounts with mostly empty fields will show absurd conversion rates, because they were usually created by hand after they converted. Backfilled records predicting the outcome that created them is the most common way a scoring model gets a flattering, useless number. In one calibration a base looked like it converted at 9.6%; on the clean cohort it was 3.2%.
When the base has no deal values, the model optimises conversations, not revenue — and will systematically favour small accounts, because small firms answer more. Say so out loud rather than presenting the ranking as a revenue ranking. Recalibrate only once deal values exist.
Hard gates beat points
Some facts should never be outweighed by a good total:
- Confirmed non-target → clamp the score below the disqualify line, no matter what else scored.
- Data doubt on a load-bearing field → clamp below the top band. The lead stays in the queue and gets human eyes, rather than being either promoted on bad data or silently dropped.
- Explicit opt-out → out of scoring entirely, out of all campaigns.
A gate that only subtracts points is not a gate. It has to clamp.
Keep the components visible
Store every block as its own column, never just the total. When a lead scores 78, you have to be able to see whether that came from a confirmed budget or from three soft engagement points — those two 78s deserve opposite actions. A single total column makes the model unauditable and, in practice, untrusted by the people meant to work the queue.
Same rule for the inputs behind a block: where a number could come from several sources, keep the sources in separate columns and note which one was used. Merging them loses the distinction between a verified figure and an inferred one exactly when you need it.
Terms and plain-language output
<!-- glossary:start terms=fit-score,lead-score-band,decision-path,deal-potential,decision-maker,day-to-day-owner,nurture,data-confidence,fit-gate,opt-out -->Rules for every document a client or their team will read (review, roast, canvas, offer cards, instruction):
- Open with «Словник цього документа»: only the terms that actually appear in it, at most 10, each as «термін: що це простими словами». Like the definitions section of a contract.
- No internal codes or letter scales in the text (A/B/C, D L T E, BLEND numbers without words, Own/Shared/Generic, Z1.4, GATE 2.5, T1–T6). Write the meaning instead: «є цифра з клієнтом», not «A».
- Use the plain form from the table below. English stays only for tool names and for copy that ships in English (promises, hooks, subject lines). If the document is in another language, translate the plain form, never the internal term.
- Do not score what cannot exist yet. A block that depends on work not done (signal catalog, case studies, a previous step) gets «пропущено: спершу потрібно <що саме>», not a score.
- Every section of a review opens with one concrete action («до наступного дзвінка: …»), then at most five lines of why. No paragraph that only restates the framework.
| Internal term (this skill) | Write for the reader | What it means |
|---|---|---|
| ICP fit, services fit, prospect score, prospect profile | оцінка відповідності профілю | Бали за те, чи компанія з вашої ніші і чи їй бракує саме вашої послуги; рахуємо ще до першого контакту. |
| band (60+ / 40–59 / below 40) | група за сумою балів | оцінку ліда переводимо в групу (наприклад, гарячий · теплий · холодний), і від групи залежить наступний крок |
| decision path | як ухвалюють рішення | Хто підписує і через які кроки проходить покупка в цій компанії. |
| confirmed deal potential, deal size | підтверджений бюджет | Сума, яку назвав сам клієнт; число, яке ми припустили, сюди не йде. |
| decision maker | ОПР (особа, що приймає рішення) | той, хто підписує: власник, операційний директор, керівник напряму |
| day-to-day owner, champion | той, хто працюватиме з вами щодня | керівник проєкту, BIM-менеджер, керівник розробки; йому важливе інше, ніж ОПР |
| nurture | догрів | регулярні корисні дотики до тих, хто ще не готовий купувати. |
| data confidence H / M / L | впевненість у даних: висока · середня · низька | Наскільки заповнені й перевірені поля компанії; від цього залежить, чи можна довіряти оцінці. |
| gate, fit gate, floor, hard gate, clamp | стоп-фільтр (мінімальний поріг) | Межа, нижче якої компанії не пишемо, хоч би що ще про неї було відомо. |
| opt-out, hard no | просить більше не писати | Людину одразу прибираємо з усіх кампаній, без спроб переконати. |
Output
One row per lead: Name · Company · Prospect profile · Engagement · Deal · Decision path · Timeline · Access · Total · Band · Action · Data confidence · Gate fired (if any) · What is missing.
Sorted by band, then by what is missing — because the top of a "needs more information" list is usually a better use of an hour than the bottom of the qualified one.
Handoff
Band → action: a call, one qualifying reply, a nurture track, or disqualify. The reply itself is
written by reply-objection-handler, which also triages it for intent and speed. Patterns across
many scored leads — a whole band that never converts — belong in reply-audit and then back into
hypothesis-builder.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/victor-shulga/outbound-engine-skills/lead-scoring">View lead-scoring on skillZs</a>