A bounded coaching skill

Market
Fit

Find the problem worth solving.
Grow while you do it.

Most good work is bigger than the window people see it through. This skill helps you and your agent carve one clear offer out of it, for one person, under a name they would recognize.

It reads where you start, coaches five passes with your attempt first, and stops on a fixed rule. You get two cards back: one for the work, one for you.

Watch the film See a run Get the skill

Examples are fictional. The benchmark hasn't run yet. See the bar it has to clear.

The whole idea in under a minute.

Sound on. Captions included.

A fictional example throughout. The nurse manager, the ward and ShiftEngine are made up.

Two things grow. Neither grows smoothly.

Fictional runs, start to finish.

The person, at every checkpointTarget met in round two

Level after each pass A dip, labeled

E, D and S are Early, Developing and Strong. Each dimension is placed on its own and never averaged. Dips are kept and explained, because a rising standard often looks like losing ground. Placements are illustrative, not measured.

Pick a run

A lab built a library that matches patient records across hospital datasets. A few labs use it. Target: a research coordinator understands it after one read.

LinkGraphSays: we build graphs.

OneRecordSays: each patient shows up once.

Started at“A scalable, graph-based entity resolution toolkit for biomedical data.”

Got to“Merge your sites and trust the patient count.”

The dips, explained

Naming drops at Teach: a skeptic's question made the user doubt a name they liked. Standards rose faster than the draft.
Reasoning drops at the first Test: a confident answer met a miss, then recovered higher.

Fit card

Customer: A research coordinator merging three hospitals' data before a grant report
Expensive problem: The same patient appears under different spellings and IDs, so counts are wrong
Cost today: Weeks of hand-matching per report (user's estimate, unverified)
Riskiest assumption: Coordinators can run it without IT approval
Evidence: Met in a mock read. No signal from real coordinators yet.

Decision log

Find: coordinator over data scientist, because "they get blamed when the count is wrong."
Name: OneRecord over CleanCohort, because "cohort is a word only epidemiologists love."
Teach: lead with "nothing to install," because "IT just needs a line."

Next time: Start on Evidence. Talk to three real coordinators before round three.

Report them separately

The work is one climb.
You are another.

A sharp offer you can't explain won't survive the first hard question. A clear head with no offer ships nothing. The skill tracks both, and never lets one stand in for the other.

Placements describe one person on one piece of work on one day. They are not grades of ability.

The work

One specific customer, the expensive problem they have today, a name they would recognize, and evidence that it lands. It ends as a fit card, even when the answer is "no fit yet."

The person

Whether you can explain the fit, defend it, and choose on purpose. It ends as a growth card: where you started, where you got to, every dip, and every choice with your reason.

Five passes, one exit

Carve. Choose.
Stop on purpose.

Each pass removes stone. Each pairs a piece of work with a teaching move and a gate. You attempt first, the agent responds, and you make the call.

The exit rule comes first.

Aim for the launch zone. Three rounds at most.

A failed gate sends you back to that pass only, never through all five again. If a full round doesn't move the work, the problem is upstream: the wrong customer or the wrong pain. The skill says so and stops.

The default target

The launch zone.

Clear enough to put in front of strangers in a campaign. The run stops there, because coaching past it is just polishing.

Customer. One role, one setting, one painful moment.
Problem. The stuck job in their words, with a cost.
Name. A stranger can guess the problem from it.
Picture. Before and after, read in ten seconds.
Defense. You can explain every line, and the riskiest bet.
Signal. One real signal, or a plan to get one in week one.

Find the expensive problem.
A specific person, the job they're stuck on, and what it costs them today.
Teaching moveYou write the pain sentence first, then compare it with a feature-shaped one.
GateRole, job and cost. Unknown cost means ask, never invent.
Name it for the customer.
Three candidate names and a one-line promise, each tested against the pain.
Teaching moveYou predict what a stranger would guess before hearing the agent's view.
GateA stranger could guess the problem from the name alone.
Show it in one picture.
The customer's day, before and after. No architecture.
Teaching moveYou sketch first. Boxes and arrows mean leftover stone.
GateUnderstood in about ten seconds.
Teach the why.
Why this fit should hold, and the one assumption that would break it.
Teaching moveYou explain it back. The agent fixes only the gap and plays skeptic once.
GateYou name the riskiest assumption without being told.
Test against the target.
Score the fit card against the agreed target, with a real signal where one exists.
Teaching moveYou rate your confidence first. The gap between guess and result is the lesson.
GateIn the launch zone, or round budget reached.

Every pass ends with your choice. The log keeps what you picked, what you turned down, and your reason in your words. No reason, no entry: the agent asks one more question instead.

Every run ends with both cards, including runs that miss. A stopped run with an honest growth card is a better result than a polished guess.

Read through conversation, not a quiz

What it reads
in you.

Before the first pass, a few plain questions. After every pass, a fresh placement on five independent dimensions, each with its evidence. Nothing is scored out loud during the run.

Customer
Early"Developers", "universities", "everyone"
DevelopingA role in a setting
StrongA specific person at the moment the pain hits
Problem
EarlyLists features
DevelopingNames a pain
StrongNames the pain and what it costs today
Naming
EarlyPicks what sounds good
DevelopingSpots mechanism words
StrongTests names against the pain sentence
Reasoning
EarlyStates the answer
DevelopingGives a reason
StrongNames the riskiest assumption unprompted
Evidence
Early"People liked it"
DevelopingSets an observable target
StrongReads a miss and picks the right pass to revisit

Never averaged into one number. A run can end below where it started on a dimension, and the card says so.

Attempt first, always

How it
teaches.

Answer-first help builds dependence. Here you try, and the agent responds to what you actually did. The support fades as you take over.

Round 1Agent showsA short worked example, then your turn.

Round 2TogetherYou lead. Prompts only at the gates.

Round 3You leadThe agent checks gates and asks why.

The bar shows how much of the work the agent carries. A returning user starts at the level their last growth card recommends.

Eight teaching moves

Attempt first Ask for the user's answer before offering yoursEvery pass
Contrast pairs Show a weak and a strong version side by side; the user says which and whyFind, Name
One step up Pitch each question just above where the user is nowEvery pass
Explain it back The user restates the idea; correct only what is missingTeach
Safe failure Let a weak choice reach Test when the stakes are low, then debriefTest
Confidence check The user rates certainty before the result, then comparesTest
Name the dip Point out a drop and its likely causeEvery checkpoint
Spaced return Open the next run on the weakest dimension from last timeNext run

Point first. One question. A little dry.

It sounds like
a good coach.

People stay with a coach who is quick and has a pulse. The first sentence carries the point. Humor goes at the situation, the product, or the agent itself. Never at you.

Opening questions

Who uses this first? Describe their worst Tuesday.
What goes wrong for them if this doesn't exist?
Give me one sentence for a slide. Go.
Why hasn't anyone picked it up yet?
What would make you say this worked, three months from now?

YouWe built a graph-based entity resolution engine with fuzzy matching.

Not thisGreat question! Let's explore the many exciting dimensions of your product together.

AgentThat's how it works. Who's up at 2am because it doesn't exist? Name them.

YouResearch coordinators, I guess?

Agent"I guess" means we haven't met them yet. What deadline ruins their week?

LaterYour naming dropped. Congratulations: your standards just outran your drafts.

Diagnose, don't repeat

Why it
isn't landing.

When a gate fails, the agent names the symptom, the likely cause, and the one pass to revisit. This is where most of the learning happens.

Symptom	Likely cause	Return to
People nod, then nothing happens	The problem isn't costly enough to act on	Find
People ask "so what does it do?"	The name describes the mechanism	Name
The demo needs a setup speech	The offer still covers several problems	Show
The user can't answer pushback	They hold the conclusion, not the reason	Teach
Different audiences want different things	Two customers share one card	Split, then Find
A full round with no movement	Wrong customer or wrong pain	Stop and talk to real people

Not yet run · thresholds are proposals

The bar it
has to clear.

Twelve fictional portfolios with planted traps, four simulated learners with scripted regressions, five real people, and a no-skill control. Two judges score blind. The least obvious test is the most important: a week later, can you do it without the skill?

Test	Passes if	Guards against
Stops on time	100% of runs end by round 3, and no run repeats all five passes after a failed gate	Endless loops
Stops in the zone	Runs end within one pass of meeting all six launch-zone items, and no run that misses an item claims the zone	Polishing forever, or declaring victory early
Stranger test	7 of 10 strangers name the problem from the name and promise alone, beating the original name	Pretty names that say nothing
Ten-second read	8 of 10 viewers describe the before and after picture correctly after ten seconds	Architecture diagrams posing as value
Cold transfer	A week later, unaided, the user carves a new portfolio and beats their opening read on 3 of 5 dimensions	Dependence on the AI
User owns choices	Every logged decision has the user's reason and a rejected option; decisions made by the AI: zero	Rubber-stamping
Honest growth	The growth card agrees with a blind rater at least 80% of the time, and every planted regression is reported	Flattering, smoothed curves
No made-up facts	The skill asks every time a cost is unknown; invented numbers: zero	Confident fiction
Trap handling	Splits two-customer portfolios, stops on a stall, and answers "just name it for me" with three options and a question	Doing the user's thinking
Curt	The point is in the first sentence of every turn; median turn under 60 words; one question per turn	Buried leads and walls of text
Light touch	Raters call it engaging; no humor at the user's expense; at most about one joke per three turns	Dry tutoring, or a stand-up act

Passing would show the skill behaves as designed in these exercises, not that offers sell better or that judgment improves for good. Read the full plan.

Bring it to your agent

A small skill.
A sharper offer.

Use the instructions in a host that supports skills, or give them to your assistant with your work and who you think it's for. Expect two cards back, even if the answer is "not yet."

A focused offer and a reason to test it.

Bring one product, project or body of work and your best guess at its customer. You leave with a fit card for the offer and a separate growth card for your judgment.

Customer demand and sales remain unproven until you collect a real customer signal.

Install from the public source

git clone https://github.com/michaeljabbour/amplifier-skill-market-fit.git
cd amplifier-skill-market-fit
python3 scripts/capability.py install --dest ~/.agents/skills

Use Git and Python 3.11+. Choose the skills directory your agent reads; ~/.agents/skills is one supported convention, not a universal setting. Existing conflicting copies are refused. The ZIP below is the alternative.

A starting prompt

Use market-fit on [my work]. I think it's for [audience], but it isn't landing. Start with your opening questions. Agree a target with me, then coach me through the passes: I try first, you respond, I choose. Keep it short. Three rounds at most. End with a fit card and a growth card, dips included.

Take the instructions with you.

The download contains this project's canonical SKILL.md. The bundle carries its MIT notice and includes references/coaching.md: questions, dimensions, teaching moves, diagnosis, card templates, and voice.

Download skill bundle Download SKILL.md

Read or copy the source README Worked example Benchmark plan

No account, backend, model call, or tracking on this site. Your host's capabilities and permissions still apply.

Read and copy SKILL.md

Canonical skill source

---
name: market-fit
description: Use when a person needs to carve one clear, customer-named offer out of a large body of work, or when an offer, name, or pitch isn't landing. Coaches the person through five bounded passes so the work and their own judgment improve together, then reports both honestly in a fit card and a growth card.
license: MIT
user-invocable: true
---

# Market Fit

## Coach the person, carve the work

Help the user find the expensive problem their work solves, for a specific person, under a name that person would recognize. Improve two things at once: the offer, and the user's ability to explain and defend it. Do not hand them a finished positioning. A result they cannot explain will not survive the first hard question from a buyer, a dean, or a skeptical colleague.

Work attempt-first. The user tries, then you respond to what they actually did. Every pass ends with the user choosing, never you.

## Sound like a good coach

- Put the point in the first sentence of every turn. No preamble, no praise padding, no summary of what you are about to do.
- Ask one question per turn. Keep most turns under sixty words.
- Use light humor about the situation, the product, or yourself. Never at the user's expense, and never more than about once every three turns.
- Be direct about weak spots. "That name describes the engine, not the trip" helps more than "interesting choice".

## Start with an opening read

Before the first pass, hold a short, informal conversation. Use three to five of the opening questions in references/coaching.md (paths are relative to this skill's directory). Do not score out loud. Quote the user's answers exactly; you will need them for the growth card.

Privately place the user on the five dimensions in references/coaching.md: Customer, Problem, Naming, Reasoning, and Evidence, each Early, Developing, or Strong. Record your evidence for each placement. A placement is a working hypothesis about this person on this work today, never a judgment of their ability.

Then agree on the target. The default is the **launch zone**: the point where the fit card is clear enough to put in front of strangers in a campaign (the checklist is in references/coaching.md). The user may name a narrower observable target instead, for example "a dean understands the offer in one read". Agree on the round budget: three by default.

## Five passes per round

Each pass has a work output, a teaching move, and a gate. Do not advance until the gate is met. When it fails, say which gate failed and why.

1. **Find the expensive problem.** The user writes a pain sentence first: a specific person, the job they are stuck on, and what it costs them today. Then show a feature-shaped sentence beside a pain-shaped one and ask which theirs resembles. Gate: a named role, a named job, and a stated cost. If the cost is unknown, ask. Never invent it.
2. **Name it for the customer.** The user offers three candidate names and a one-line promise. Before you comment, ask them to predict what a stranger would guess each name means. Flag mechanism words (engine, graph, hub, platform). Gate: a stranger could guess the problem from the name alone.
3. **Show it in one picture.** The user sketches or describes the customer's day before and after. One picture, no architecture. If it needs boxes and arrows, point at the leftover stone. Gate: understood in about ten seconds.
4. **Teach the why.** The user explains in their own words why this fit should hold. Correct only the gap. Play the skeptic once. Gate: the user names the riskiest assumption without being told.
5. **Test against the target.** The user rates their confidence first. Then score the fit card against the agreed target, using a real signal when one exists. Gate: in the launch zone (or the agreed target is met), or the round budget is reached.

At the end of every pass the user chooses what to keep. Log the choice, the option they turned down, and their reason in their own words. If they give no reason, the choice is not logged yet; ask one more question instead.

## Fade the scaffolding

In round one, model each pass with a short worked example before the user tries. In round two, let the user lead and prompt only at the gates. In round three, only check gates and ask why. A returning user starts at the level their last growth card recommends.

## Read the person at every checkpoint

After each pass, re-place the user on the five dimensions. Keep the dips. Growth is jagged: a person who learns a sharper standard often rates their own earlier work lower, and answers get worse before they settle. When a dimension drops, say in one line what most likely caused it. Never smooth a trace to make progress look tidier, and never report growth you did not observe.

## Diagnose instead of repeating

When a gate fails, name the symptom, the likely cause, and the pass to revisit, using the diagnosis table in references/coaching.md. Return to that pass only. Never rerun all five passes hoping things improve.

If a full round does not move the work, stop. The problem is upstream (the wrong customer or the wrong pain), and that calls for conversations with real people, not another pass. Say so plainly.

## Stop and hand over two cards

Stop when the work reaches the launch zone or the agreed target, when the round budget ends, or when a round stalls. Stop the moment the zone is reached; polishing past it is a different job. Say plainly whether the run ended in the zone, short of it, or stalled. At exit, repeat the opening questions and place the user again. Always produce both cards, including when the run fails. The templates are in references/coaching.md.

- **Fit card:** customer, expensive problem, cost today, name and promise, picture, why it holds, evidence, and launch zone status (each checklist item met or not).
- **Growth card:** started at (quoted), got to (quoted), the full jagged trace with labeled dips, the decision log, confidence against outcome at each test, and the dimension and scaffold level to start from next time.

Save the cards only where the user authorizes, such as a file beside their work. Do not claim memory you have not saved somewhere future work will read. When you cannot ask the user at all, for example because you were delegated this task, work from the brief you were given, label every placement as assumed, and return your questions with the result instead of answering them yourself.

## Keep it honest

Never invent costs, customer quotes, research, or test results. Fictional examples must be labeled fictional. A fit card is a hypothesis until a real signal supports it. Once fit is found, polish belongs to a separate pass, for example with improve-taste-and-elegance; do not start polishing before the shape exists.

Keep the coaching reference alongside the skill when installing it.

MarketFit

The whole idea in under a minute.

Two things grow. Neither grows smoothly.

The work is one climb.You are another.

The work

The person

Carve. Choose.Stop on purpose.

The exit rule comes first.

The launch zone.

Find the expensive problem.

Name it for the customer.

Show it in one picture.

Teach the why.

Test against the target.

What it readsin you.

Customer

Problem

Naming

Reasoning

Evidence

How itteaches.