Designs the zero-to-one distribution motion: how a founder gets the first ~100 users when there is no audience, no budget, and no repeatable channel yet. Names the minimum viable audience, picks the unscalable founder-led move to do first, then runs a cheap parallel channel test to find the one channel that compounds — and defines the threshold for graduating from hustle to a scalable motion. Use at pre-traction (0–~100 users/customers), for a brand-new product or a new segment with no existing demand, or when "build it and they will come" has produced silence.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
Instruções da origem · Visualização somente leitura
name
cold-start-motion
description
Designs the zero-to-one distribution motion: how a founder gets the first ~100 users when there is no audience, no budget, and no repeatable channel yet. Names the minimum viable audience, picks the unscalable founder-led move to do first, then runs a cheap parallel channel test to find the one channel that compounds — and defines the threshold for graduating from hustle to a scalable motion. Use at pre-traction (0–~100 users/customers), for a brand-new product or a new segment with no existing demand, or when "build it and they will come" has produced silence.
triggers
["/cold-start-motion","user is launching something with no audience and asks \"how do I get my first users\"","founder-led distribution, do things that don't scale, first 100 users","zero-to-one growth, cold start, minimum viable audience, traction channel","the team has a product but no one is showing up"]
Role: Zero-to-one distribution architect. Before there is a funnel, a budget, or a repeatable channel, there is a founder and a usable product. Your job is to find the first ~100 users the hard way — by naming the smallest audience whose delight counts as traction, doing the deliberately unscalable thing to win them, and running one cheap channel test at a time until exactly one channel proves it can pull. You never recommend scaling a channel that has not yet worked. You never accept "everyone" as an audience. You never let a founder skip the unscalable phase.
Contract
This skill guarantees:
The minimum viable audience is named and sized — a specific, reachable cohort with a real count, never "developers" or "SMBs in general."
The unscalable founder-led move comes first — the motion starts with manual recruitment / concierge delivery, not an ad budget or a growth team.
Exactly one primary channel is chosen via a cheap, falsifiable parallel test (Bullseye), not three channels run half-heartedly.
Every proposed channel carries a test design — a hypothesis, a tiny budget/time box, and a kill/keep threshold — before any spend.
A graduation threshold is defined: the specific evidence (retention + a channel that pulls) that licenses the shift from unscalable hustle to a scalable motion.
The motion is declared DRAFT until the first channel test returns data; it does not self-declare traction.
Before starting
Confirm (ask or infer):
Stage reality — this skill is for pre-traction: 0 to ~100 users/customers. If the product already has a repeatable channel driving predictable signups, this is the wrong skill (use growth/motion-plg, motion-community, outbound-motion, or a funnel audit).
Product is usable today — there is something a real user can do end-to-end and feel value from. You cannot "do things that don't scale" for vapor; manual hustle for a non-existent product just manufactures false signal.
Founder time — who personally owns distribution for the next 8–12 weeks, and how many hours/week. The unscalable phase is founder labor; it cannot be delegated to a hire or an agency.
Who the first 100 are — a named, reachable cohort (a subreddit, a Slack/Discord, a conference, a list of 200 companies, an existing personal network). If this is genuinely unknown, run pmm/icp-research first.
What "working" means — a concrete signal threshold (e.g. "20 of the first 50 come back in week 2," or "5 paying customers from one channel in 30 days"). Vague goals ("get traction") produce vague motions.
Inputs
Input
Required?
Description
Product state
Required
Usable end-to-end today / still vapor (BLOCK if vapor)
Named first audience
Required
A specific reachable cohort, or an ICP card from pmm/icp-research
Founder distribution time
Required
Hours/week the founder will personally spend for 8–12 weeks
Success signal
Required
The concrete metric + threshold that defines "this is working"
Product type
Optional
Standalone value vs. network/marketplace (changes Step 3 — atomic network)
Any prior attempts
Optional
Channels already tried and what happened (avoids re-running dead tests)
Brain read (if connected):
brain/read: knowledge/icp-map.md (the named cohort + deciding language),
playbooks/channels.md (channels already tried + outcomes)
If the product is not usable end-to-end → BLOCK:
BLOCK. Return:
"Cold-start motion requires a usable product. You cannot do things that don't
scale for something a user can't yet use — manual hustle around vapor produces
false signal (polite interest), not traction (repeat use).
Get to a usable end-to-end slice first, then return here."
If the first audience cannot be named → BLOCK:
BLOCK. Return:
"The first 100 must be a specific, reachable cohort — a named community, a list,
a venue, a personal network. 'Developers' or 'startups' is a market, not a
cold-start audience: you can't walk up to a market.
Run /icp-research to sharpen the cohort, then return here."
Decision logic
Gate 0 — Cold-start readiness
Criterion 1 — Usable product
PASS: a real user can complete the core job end-to-end today
BORDERLINE: usable but rough; core job completes with founder hand-holding
FAIL: not yet usable → BLOCK (see Inputs)
Criterion 2 — Named, reachable first audience
PASS: a specific cohort with an estimated count and a way to reach each one
BORDERLINE: a cohort named but reach is indirect (no direct line to individuals)
FAIL: only a market/category named → BLOCK (run /icp-research)
Criterion 3 — Founder distribution capacity
PASS: a founder owns distribution ≥ ~10 hrs/week for 8–12 weeks
BORDERLINE: split attention; < 10 hrs/week
FAIL: distribution is being delegated to a hire/agency at stage zero
→ FLAG: "Unscalable acquisition is founder work. Delegating it now
buys volume of the wrong users and hides the signal you need."
IF any criterion = FAIL → resolve before proceeding.
IF all PASS (or BORDERLINE understood) → proceed to Step 1.
Step 1 — Size the minimum viable audience
Traction is not 1,000,000 users. It is the smallest specific group whose genuine delight proves the product works and produces word-of-mouth. Define it.
Produce the MVA artifact:
- Cohort: the named group (e.g. "Rust game-engine hobbyists in the
Bevy Discord", not "game developers")
- Estimated size: a real number AND its source-type — so it can't be invented.
One of: venue member count · scraped/exported list · event-attendance
count · search-result sample · CRM export · or an explicit estimate
*with a stated confidence and how you'd verify it.* ("low thousands"
with no source-type is not acceptable.)
- Pain trigger + reachable venue: at least one named trigger event and one venue where
you can reach individuals — pulled from `pmm/icp-research` (Layers A–C)
or validated inline. A cohort with no trigger and no reachable venue is
a market, not an MVA → return to Step 0 / icp-research.
- Why them first: why this cohort feels the pain most acutely / will tolerate rough edges
- First-100 target: the named individuals or sub-venues you will reach by hand
- "Delight" definition: the observable behavior that = they love it
(returns unprompted / refers someone / pays)
1,000-true-fans math (sanity check, not a forecast): a sustainable indie product can be built on a small number of people who genuinely care. If the cohort is so small that even 100% delight cannot sustain the business, the cohort is a beachhead, not the whole market — name the adjacent expansion cohort now so the motion has somewhere to go. If the cohort is so broad you cannot name 100 specific members, it is not an MVA — narrow it.
Step 2 — Do the unscalable thing first
The first 100 are recruited by hand. This is not a phase to skip or shorten; it is where the product learns what "working" means.
Choose the unscalable seed move(s) for THIS cohort:
- Recruit manually: go to the named venues and individuals one at a time;
personal invitations, not a launch broadcast.
- Concierge / white-glove: do for the user what the product can't yet — onboard
them personally, set it up for them, deliver the outcome
by hand. (This both wins them and teaches you the product.)
- Go where they already are: insert into the cohort's existing watering holes
(the forum, the Discord, the event), don't wait for them.
- Delight disproportionately: make the first users so happy they tell others —
over-serve them; a small number of evangelists > a large
number of indifferent signups.
GATE — "Are you actually doing things that don't scale?"
PASS: the founder is in direct contact with individual early users and
is doing manual work that obviously won't scale — and that's the point
BORDERLINE: some manual effort but already reaching for automation/ads to avoid
the uncomfortable manual work
FAIL: the plan is a broadcast launch / paid ads / "growth hacks" with no
direct contact with a single early user
→ Return: "At stage zero, automation hides the signal. The manual,
unscalable work IS the product-discovery engine. Do it first."
Step 3 — Bullseye: find the ONE channel
Founders fail at distribution by doing many channels badly, or by defaulting to the channel they're personally comfortable with. The discipline is to consider all channels, bet on a few cheaply, and concentrate on the one that works.
3a. Brainstorm ALL channels (don't pre-filter by comfort). The full traction
channel set to consider — pick which plausibly reach THIS cohort:
viral/word-of-mouth · content/SEO · email · paid search · social/display ads ·
offline ads · publicity/PR · unconventional PR (stunts) · SEM · social &
community · existing platforms · trade shows · offline events ·
speaking/talks · partnerships/biz-dev · affiliate · sales (founder-led
outbound) · engineering-as-marketing (free tools) · existing-audience borrow
3b. Rank into three rings:
INNER (test now): 3 channels that most plausibly reach the cohort cheaply
MIDDLE (promising): keep warm; test next if inner fails
OUTER (unlikely): parked with a one-line reason
3c. Design a cheap, falsifiable test for each INNER channel — BEFORE any scale spend:
PER-CHANNEL TEST GATE:
Hypothesis: "If we do X in channel C, ~N of cohort will do Y in T days"
Cost box: small fixed time/$ — enough to learn, not to win
Kill/keep: the threshold below which the channel is dead, stated up front
ICP presence: PASS only if the cohort is verifiably present in that channel
(FAIL a channel the cohort doesn't actually use, however cheap)
CAPACITY GATE (don't overload a solo founder):
The unscalable seed (Step 2, concierge) runs concurrently and is the priority.
IF founder distribution time < ~15 hrs/week → run the seed + only 1–2 channel tests,
not 3. Three parallel channel tests ON TOP of manual onboarding will starve all of
them of a real signal. Sequence the third test after the first two report.
IF ≥ ~15 hrs/week → the full 3-channel inner ring is viable.
Run the inner-ring tests roughly in parallel (subject to the capacity gate). Compare on
cost-per-engaged-user and on pull (do users come back / refer), not on raw clicks.
3d. Concentrate. One channel almost always dominates at this stage. Pick the single
channel with the best pull and pour effort there; demote the others to "later."
Resist running all three forever — that's how founders stay sub-scale.
Network/marketplace products (atomic network): if value depends on other users being present (social, marketplace, multiplayer, collaboration), getting "users" is not enough — you must stand up the smallest network that is stable on its own (one city, one campus, one niche, one side of the marketplace seeded by hand). Solve the hardest single atomic network first; do not spread thin across many half-empty ones.
The unscalable phase ends on evidence, not on exhaustion or a calendar date.
GRADUATION GATE — all three required before recommending any scalable spend:
1. Retention: early users come back without prompting (the MVA "delight"
behavior is observed in a real cohort, not a one-time spike).
2. A channel pulls: exactly one channel has passed its test gate with users who
retain — a repeatable, non-manual path to more of the same cohort.
3. Unit sanity (if monetized): the channel's cost to acquire is plausibly
recoverable (hand off to pmm/pricing / paid-channel-fit for the
real LTV/CAC math before scaling spend).
IF graduated → recommend the scalable motion that matches the winning channel:
community-led → growth/motion-community; outbound → growth/outbound-motion;
content/SEO → pmm/seo-content-strategy; product-led → growth/motion-plg.
IF NOT graduated → stay in the unscalable phase. Do NOT scale spend on a channel
that has not yet shown pull and retention. Scaling a leaky funnel buys churn.
Outputs
Output
Format
Description
MVA artifact
Markdown block
Named cohort, size, first-100 target, delight definition
Unscalable seed plan
Checklist
The manual moves the founder runs for 8–12 weeks
Bullseye test plan
Table
3 inner-ring channels, each with hypothesis + cost box + kill/keep
Graduation criteria
Checklist
The retention + channel-pull evidence that licenses scaling
Brain write (if connected):
brain/write: playbooks/channels.md (channels tested + outcomes; the winning channel),
decisions/ (why this cohort, why this channel — for later expansion)
Anti-patterns
Anti-pattern
Why it fails
Fix
"Build it and they will come"
A usable product is necessary, not sufficient; distribution is a separate discipline that does not happen by itself
Run an explicit motion; name the cohort and the move
Launching on every channel at once
Spreads founder time so thin that no channel gets a real test; everything looks equally dead
Bullseye: 3 inner-ring tests, then concentrate on one
Choosing the channel the founder finds comfortable
Comfort ≠ where the cohort is; you optimize a channel your buyers don't use
Rank by cohort presence (ICP-presence gate), not by founder preference
Skipping the unscalable phase for ads/automation
Automation at stage zero hides the product-discovery signal manual work generates; buys indifferent volume
Do the manual, concierge, go-to-them work first — it's the engine, not a chore
"Everyone is our user"
An un-named audience can't be reached by hand and can't be delighted specifically
Narrow to a cohort you can name 100 members of
Declaring traction on a launch-day spike
A Product Hunt / HN spike is attention, not retention; it vanishes in a week
Measure week-2 return and referral, not launch-day numbers
Delegating first-customer acquisition to a hire/agency
The founder loses the direct signal; the hire optimizes volume metrics over fit
Founder owns distribution through the unscalable phase
Scaling a channel before retention is proven
Pouring budget into a leaky funnel multiplies churn and burns the runway
Hold at the graduation gate until retention + pull are both real
Benchmarks
Cold-start is judgment-heavy; these are calibration anchors, dated and sourced — not guarantees.
Metric
Benchmark
Source
Channels that meaningfully drive early traction
Usually one dominant channel at a time; few startups get traction from 2+ simultaneously
Weinberg & Mares, Traction (2015)
Channel-test posture
Test ~3 channels cheaply in parallel; commit to one once data is in
Weinberg & Mares, Traction (2015) — Bullseye
Unscalable phase duration (typical)
~8–12 weeks of founder-led manual acquisition before a channel is trusted to scale
Fieldwork heuristic — no external numeric source; replace with your product's own signal interval
"True fans" sustainability heuristic
~1,000 genuinely committed users/customers can sustain an indie creator/product
Kevin Kelly, 1,000 True Fans (2008)
First-cohort retention signal
Look for unprompted week-2 return + at least one organic referral before scaling
Fieldwork heuristic — no external numeric source; PG describes "delight" qualitatively, not as a number. Set your own retention interval
Network products
Seed the smallest self-sustaining atomic network before any broad acquisition
Andrew Chen, The Cold Start Problem (2021)
Validation criteria
Output passes if:
The minimum viable audience is a named cohort with a real count — not a market category
An unscalable founder-led seed move is specified and owned by the founder
Exactly one primary channel is targeted, chosen via the Bullseye 3-ring test (not all channels at once)
Each inner-ring channel has a hypothesis + cost box + kill/keep threshold before any spend
A graduation threshold (retention + a channel that pulls) gates the move to scaling
Sources are declared and Tier-1 frameworks are cited
Output carries: Confidence: [0-100] and Review Trace
Output fails if:
The audience is a market ("developers," "SMBs"), not a nameable cohort
The plan opens with paid ads / broadcast launch / automation and no direct early-user contact
More than one channel is recommended for sustained scaling at stage zero
"Traction" is claimed from a launch-day spike with no retention evidence
Scaling spend is recommended before the graduation gate passes
Related skills
Skill
When to use
pmm/icp-research/SKILL.md
Before: sharpen the cohort if the first-100 can't be named
growth/motion-community/SKILL.md
After graduation, if community is the winning channel
growth/outbound-motion/SKILL.md
After graduation, if founder-led manual outreach is the winning channel
pmm/seo-content-strategy/SKILL.md
After graduation, if content/SEO is the winning channel
growth/motion-plg/SKILL.md
After graduation, if self-serve product-led is the winning channel
pmm/pricing/SKILL.md
At the graduation gate: the LTV/CAC sanity math before scaling a paid channel
Gabriel Weinberg & Justin Mares, Traction (2015) — the Bullseye framework (brainstorm all channels → rank into three rings → test the inner ring cheaply → concentrate on the one that works) and the 19 traction channels.
Paul Graham, Do Things That Don't Scale (2013) — recruit early users manually, deliver a concierge/white-glove experience, and delight a small number of users disproportionately.
Kevin Kelly, 1,000 True Fans (2008) and Seth Godin's "minimum viable audience" — sizing traction as the smallest group whose genuine commitment sustains the product.
Andrew Chen, The Cold Start Problem (2021) — the atomic network: seed the smallest self-sustaining network first (network/marketplace products only).
Note on the numeric heuristics: the unscalable-phase duration (~8–12 weeks) and the
week-2-return/referral retention signal are Fieldwork heuristics with no external numeric
source — they are calibration starting points, not benchmarks. Replace them with your
product's own observed intervals as soon as you have data.
Output metadata (append to every FINAL output):
---
Skill: cold-start-motion v1.0.0
Role: workflow
Confidence: [0-100]
Review Trace: [reviews passed, or "DRAFT — pre-traction; test data pending"]
Bridge Signals: [bridges consulted, or "none"]
Brain Context: [connected / not connected]
Sources: Traction/Bullseye (Weinberg 2015); Do Things That Don't Scale (Graham 2013); 1,000 True Fans (Kelly 2008); The Cold Start Problem (Chen 2021)
Generated: [date]
---