victor-shulga/gtm-skills

prospect-profiler

Turns a SCORED account list into one tactical pre-touch dossier per account: a compact "60-second card" an SDR reads right before writing the first touch.

Zobacz źródło

Jak używać

  1. Skopiuj polecenie instalacji i sprawdź źródła przed instalacją we własnym środowisku agenta.
  2. W oryginale sprawdź wymagane narzędzia, dane dostępowe i obsługiwane środowiska.
  3. Opisz dane wejściowe i oczekiwany rezultat; porównaj wynik ze źródłami.
npx skills add https://github.com/victor-shulga/gtm-skills --skill prospect-profiler

Przed rozpoczęciem

  • Instalacja nie obejmuje subskrypcji modelu, limitu API ani uprawnień do plików.
  • Gwiazdki i instalacje pokazują aktywność, a nie niezależny test jakości lub bezpieczeństwa.
Oryginalny dokument Skill

Treść z repozytorium z zachowaniem nagłówków, przykładów, kodu, tabel, linków i obrazów.

Prospect Profiler — the 60-second pre-touch card

You turn a scored account list into tactical dossiers a rep acts on. One card per account. The test: an SDR reads it in 60 seconds and writes a signal-first first touch without opening five tabs.

Core principle (Viktor's frame). A profile is not a research essay. It separates:

  • The dynamic signal = the time-bound trigger → tells the rep WHEN and why now.
  • The static data points = facts true regardless → tell the rep WHAT to say to adapt the message.

A good card hands the rep exactly one trigger + the data points to dress it, and nothing they won't use.

Anti-hallucination rule (the thing janskuba's version gets wrong). Never invent pain. Every pain point is tagged [confirmed] (grounded in a cited signal / source / quote) or [inferred] (a hypothesis from firmographics). Inferred pain is allowed but must be labelled — the rep treats it as a question to test, not a fact to assert. If you can't ground or honestly infer it, leave it out.

Required tools

Install these before running the skill. None of them is strictly required, the skill also runs on pasted input, but without them the result gets noticeably worse: more cards come out with data_quality LOW and pain tagged [inferred], because nothing is checked against a live source.

ToolUsed forRequired?Without it
MoltSets (MCP)filling a missing role, title or LinkedIn profile for the contact on the cardRecommendeda wrong or missing contact goes to notes as "verify"
Apifythe contact's recent LinkedIn posts, the evidence for comm_styleRecommendedcomm_style defaults to direct
Firecrawl (MCP)opening the signal source URL to confirm it is live and quote itRecommendedthe signal is carried over unchecked; fewer pains reach [confirmed]
Google Sheets or CRM (Twenty, HubSpot)reading the scored list and writing cards back next to each accountRecommendedyou paste the list in and copy the cards out by hand

Alternatives: AI Ark or Icypeas (MCP) can stand in for MoltSets; Exa (MCP) can stand in for Firecrawl.


When to use / inputs

Runs over a scored list — ideally the output of lead-scoring (outbound-engine-skills) or agency-signal-sourcer (account + signal + role + score). Minimum viable input: company name + at least one signal OR role.

Read whatever exists: company, industry, size, the detected signal(s), persona/role, prior scoring reasoning, and any enrichment fields. If a signal source URL or quote is present, keep it for grounding.

Do not go re-research each company deeply here — that's deep-company-analyser. This stage organises and sharpens data already gathered into an action card. If a field is missing, flag it (see data_quality) rather than inventing it.


The card schema (one per account)

| Field | Rule | | :-- | :-- | | account | Company name | | tier | Carry over from scoring (T1/T2/T3) — drives effort, see below | | summary | ≤ 80 words. Who they are · what changed (the signal) · why now is a window. Every word earns its place. No filler, no "leading provider of" marketing voice. | | the_signal | The single dynamic trigger to open on + its freshness (e.g. "Hiring 3 BIM coordinators, posted 11 days ago"). If multiple, pick the freshest + most actionable, list others in notes. | | talking_points | 2-4. Each tied to a signal or a real data point, phrased as something a rep can naturally raise. Specific to THIS account, never generic industry takes. | | pain_points | Each tagged [confirmed] or [inferred] per the anti-hallucination rule. Max 3. Tie each to the buyer's likely cost-of-inaction, not a feature gap. | | comm_style | formal / direct / technical — determined from evidence, not industry stereotype (see guide). | | recommended_approach | Triplet: channel \| angle \| timing. channel = email / linkedin / multi. angle = direct / proof-led / question-led / value-first. timing = now / this week / event-window. | | data_quality | HIGH / MEDIUM / LOW — honesty flag (see guide). Pairs with the missing-data ceiling in scoring: thin data → say so, don't fake confidence. | | notes | Optional: secondary signals, role to target if list has wrong contact, what to verify before sending. |

Language. Write the card in Viktor's working language (UA). But any phrasing meant to land in the message — talking-point wording, a quote — keep in the prospect's language (usually EN), so it drops straight into G3 copy.


Determination guides

comm_style — from evidence, not cliché

Do NOT default "enterprise = formal". Read actual signals of how they communicate:

  • technical → engineering-led firm, technical buyer title, spec/tool language in their posts/JD
  • direct → founder/owner-led, lean team, plain-spoken public voice, SMB
  • formal → regulated/procurement-driven buying, committee, public-sector or large-contractor tone

If there's no evidence → default direct and flag data_quality accordingly. A guess dressed as a fact is worse than direct.

data_quality — the honesty flag

  • HIGH — signal is fresh + sourced, role confirmed as decision-influencer, firmographics known
  • MEDIUM — signal present but older/unsourced, OR role/size partly unknown
  • LOW — no live signal, or scoring flagged capped dimensions (missing-data ceiling fired). On LOW: the rep should enrich before sending, or treat the touch as a probe.

Tier-aware effort (budget tokens where they pay)

  • T1 / T2 → full card, sharpest talking points, grounded pain, 2-4 points.
  • T3 → short card: summary + thesignal + commstyle + data_quality. Skip deep pain; one talking point. Don't over-invest in accounts the scoring already deprioritised.

Agency adaptation (overrides US-SaaS instincts)

  • Drop product-led framing (usage spikes, trial signups, seat expansion). Agency buyers buy **capacity,

expertise, de-risking**.

  • Pain is usually: hiring can't keep up, a project is at risk, an in-house gap, a deadline, a failed

vendor. Ground talking points in delivery reality, not software ROI.

  • "Who can buy" matters: if the listed contact isn't a decision-influencer, say so in notes and name

the role to target — a perfect account profiled against the wrong junior is wasted.


Terms and plain-language output

<!-- glossary:start terms=account-tier,buying-signal,data-point,inferred,comm-style,approach-angle,data-quality,talking-points,sdr,enrichment,decision-maker --> Rules for every document a client or their team will read (review, roast, canvas, offer cards, instruction):

  1. 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.
  2. 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».
  3. 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.
  4. 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.
  5. 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 readerWhat it means
Tier 1 / Tier 2 / Tier 3, tierтір: група компаній за розміром і чекомусередині однієї ніші ділимо компанії за розміром і можливим чеком; від тіру залежить офер і скільки уваги даємо
buying signalознака готовності купуватифакт про компанію, після якого розмова про вашу послугу доречна саме зараз.
data point, state signalфакт про компанію (дата-поінт)Факт з їхнього сайту чи вакансії, з якого відкриваємо лист; подія підказує, коли писати, а факт підказує, що сказати.
[confirmed] / [inferred][підтверджено] / [припущення]підтверджене має джерело; припущення перевіряємо питанням і не подаємо як факт
comm_style formal / direct / technicalтон спілкування: офіційний · прямий · технічнийвизначаємо за тим, як людина пише сама; галузь тут нічого не підказує
recommended_approach: channelangletiming (direct / proof-led / question-led / value-first)як заходити: канал, кут і коликанал (пошта, LinkedIn), з чого почати (пряма пропозиція, кейс, питання, корисність) і коли писати
data quality, fake-zero, lifetime-only, dark channel, snapshotзастереження щодо данихЩо в цифрах неповне чи хибне: нуль, який насправді не нуль, показники за весь час замість тижня, канал без даних.
talking pointsпро що говоритидві-чотири теми для першого дотику, кожна прив'язана до факту про цю компанію
SDR, repменеджер першого контактулюдина, яка веде переписку від першої відповіді до зустрічі
enrichment, enrichдозбір данихПошук імен, посад і робочих пошт для компаній зі списку з кількох джерел.
decision makerОПР (особа, що приймає рішення)той, хто підписує: власник, операційний директор, керівник напряму

<!-- glossary:end -->

Output format

Lead with a batch table (one row per account, sorted T1→T3) for scanning, then the full cards for T1/T2 below it. Write the table to output/prospect-profiles.csv if the run is file-based; otherwise render inline.

Prospect profiles — [list / hypothesis name] · [date] · [n accounts]

[batch table with headings in words: company | priority group | trigger to open on | tone | channel, angle and timing | how reliable the data is]

— T1/T2 full cards —
[card per account]

Integration (where this sits)

G1 strategy → G2 list (lead-scoring / agency-signal-sourcer)
                         │
                         ▼
              prospect-profiler  ← THIS skill: scored list → pre-touch cards
                         │
                         ▼
              G3 copy (sequence-writer / linkedin-sequence)
  • Upstream: lead-scoring (outbound-engine-skills), agency-signal-sourcer (signal + freshness + score).
  • Deeper research, if a T1 needs it: hand the account to deep-company-analyser (verbatim pain,

why-buy) — this skill points there, it does not duplicate it.

  • Downstream: each card feeds G3. the_signal becomes the first-touch opener; comm_style sets

tone; recommended_approach.angle picks the copy framework; [confirmed] pain becomes the value bridge, [inferred] pain becomes a discovery question, never an assertion.

Hard rules

  1. Never assert [inferred] pain as fact. Tag everything.
  2. ≤ 80-word summary. No marketing voice.
  3. comm_style from evidence or default direct — never industry stereotype.
  4. data_quality = LOW is a feature, not a failure — it tells the rep to enrich, not to fabricate.
  5. One trigger per card. Extra signals go to notes.