Terminal
Terminal/Skill
Measures a company's headcount growth over several windows and buckets it from shrinking to hyper-growth, with base counts shown.
By Clay
Works with
npx skills add clay-run/clay-skill-creator --skill headcount-growthThe insight: a growth percentage is a trajectory claim built on three silent assumptions — right entity, meaningful denominator, and a window that isn't hiding a reversal — and the bare action returns a confident number when any of them is wrong. +40% over 12 months can mean a 3-person company hired one engineer, a different company than the one you asked about, or a real grower that started shrinking last quarter. And the action's miss is success-shaped AND billed: an empty result costs the same credit as a hit. This skill wraps one cheap action with the checks that make its number safe to act on.
Nothing here ships with a value. Each one is the installer's, not the author's: ask for it, never substitute a plausible default, and where an answer does not exist say which step becomes unavailable rather than guessing. Where a default IS defensible it is named below, and using it means saying so in the output.
| Input | What the installer supplies | If it is missing |
|---|---|---|
| An identifier per company | a company profile URL for the high-accuracy route, or a domain for the documented lower-accuracy one | route each row by what it has; the two run as separate calls and mixing them hard-fails. Name-only rows resolve the domain first, and if the resolution returns a profile URL, harvest it — that row then rides the accurate route for free |
| Which windows | near-term momentum, sustained trend, or both | 3 and 12 months read together is defensible and must be stated: one window alone is a number, not a trajectory |
| Cost ceiling | credits, knowing that misses bill too | dedupe companies first, state list × cost, and say that obscure and very small companies miss more — a low-coverage list burns credits on empty results |
If an answer sheet is present beside this skill, load it and ask only for what it does not cover. A partial sheet is normal; a value it is missing gets asked for on its own rather than restarting the interview. Say which values came from the sheet before using them — a sheet applied silently is a wrong field nobody catches. If there is no sheet, say nothing about sheets — the check is a file lookup, not a question, so run the interview as though the feature did not exist rather than reporting an absence. At delivery, offer to save the answers back (identifiers only — never a token or a password), private and never published — and phrase the offer so it explains itself: "want me to save your answers to a file, so the next person on your team doesn't have to answer these again?"
Run clay whoami; echo "exit_code=$?". If it fails or Clay tools are missing,
run the Clay plugin's setup skill, restart if it says to, and re-run this
skill. Tell the user which workspace you're in. Confirm the growth action in
the live catalog by DISPLAY NAME ("Find Company Headcount Growth") — action
keys drift (references/growth-mechanics.md) — and read its declared cost.
Per unique company, run Find Company Headcount Growth — one arm per row:
url = LinkedIn company URL, or website = domain (never a domain in url;
it hard-fails). Small lists (≤20): ad-hoc action execution — it has a
25-runs/day workspace quota, and if today's quota is already spent the refusal
is explicit and free — switch surfaces, don't wait. Larger lists or spent
quota: the two single-arm workflows in references/growth-mechanics.md
(workflow runs bypass the ad-hoc quota). Never loop past the quota into
errors.
employee_count + per-window backdated counts and
percentages. FIRST check the entity echo: the result's name/url name the
company the action actually matched — if they don't match the company you
asked about, the row is a wrong-entity hit; flag it, don't report its
numbers.success: true, result EMPTY (the only signal is
a "Company Not Found" preview). Verdict unverifiable — never "flat", never
0%. The credit was still spent; count it.+300% on 3→12
people ships as "grew 3→12 (micro-base)", flagged, not as hyper-growth.<0 shrinking · 0–10% flat · 10–30%
growing · 30–100% high-growth · >100% hyper-growth.reversing; growing year + a last quarter running well ahead of the year's
pace = accelerating; flat 12-month + strong 3-month = inflecting up.
Say which windows produced the verdict, and read the backdated counts for
dip-and-rebound shapes the window percentages smooth over.Per company: identity (asked → matched echo) · employee_count · per-window counts + % · bucket · trajectory · flags (micro-base, wrong-entity, window-gaps) · verdict (measured / unverifiable). Plus the roll-up: companies
in, measured, unverifiable, wrong-entity, credits measured vs declared
(misses included). Every input company lands somewhere.
name/url) on every hit; a mismatched echo is
a wrong-entity flag, never a reportable number.unverifiable (billed, counted) — never
zero growth; MUST treat per-window nulls as missing snapshots, never 0%.Ask: "Which of these 30 accounts are actually growing? CSV has name, domain,
some LinkedIn URLs." Scope: dedupe 30 → 28; 9 rows LinkedIn+domain, 16 domain
only, 3 name only → resolved first (1 ambiguous, parked). Approval at ~28
credits. Run: workflow surface (batch > 20). Read: 24 hits (entity echo clean
on 23 — 1 mismatch flagged: a same-named company matched from a bare domain),
3 misses → unverifiable (credits counted), plus the parked ambiguous row.
Interpret: 6 high-growth (12-mo), but 2 of them show negative 3-month deltas →
stalling flag; 1 shows +180% on a 14-person base → micro-base flag, not
hyper-growth. Deliver: 23 measured rows with counts + buckets + trajectories,
5 unmeasured (3 unverifiable, 1 wrong-entity, 1 unresolved), 28 credits
measured vs 28 declared.