Source: frameworks/kit-micromagnet/01-micromagnet-context.md

MicroMagnet Factory — Context & Inputs

Config (edit as business pressure shifts)

Offer names and URLs (used in the CTA PowerUp block of every shipped micromagnet):

Low-ticket offer name: [SET ME]
Low-ticket offer URL: [SET ME]
$27 workshop name: [SET ME]
$27 workshop URL: [SET ME]
Recurring offer name: [SET ME]
Recurring offer URL: [SET ME]

Voice anchor files (reference for tone, sentence length, and register — don't reproduce their structure, match their voice):

content/business/marketing/campaigns/practice-command-center/01-client-intelligence-brief-skill.md
content/business/marketing/campaigns/practice-command-center/02-hidden-revenue-scan-skill.md

CTA destination priority tilt: low-ticket paid offer

The tilt names which destination the micromagnet's CTA leans toward when multiple destinations fit. The Kit still picks per-micromagnet based on what the Gap logically points to — the tilt governs only close calls. Edit as the business shifts.

If any Config value needed at the CTA step is still [SET ME], stop and flag before producing the micromagnet. Don't ship with placeholder URLs.

What You Need Before Building

Input	Where It Comes From	Required?
Micro-problem	Generated (Step 0a or 0b) or named by Kathryn	Yes
Pre-AI precedent	Research — what did people pay for before AI that solved this?	Yes
Output description	Design — what does the user walk away with? (Not what the skill does. What they hold in their hands.)	Yes
Input path	Design — what does the user paste? Must be something they already have.	Yes
Target audience	Reference — `business-aos/reference/core/audience.md` or the relevant offer brief	Yes
Cannibalization check	Review — does this overlap with any paid offer?	Yes
The Gap	Design — what adjacent problem does this deliberately leave unsolved?	Yes
Positioning decision	Niche / Universal / Both	Yes
Nano spec	Which step of the cheat code gets extracted standalone	Yes

Step 0: Two Intake Paths

Run one of 0a or 0b depending on where the source material lives. Skip both if Kathryn names the micro-problem directly.

Step 0a: Audience-theory intake

Use when starting from strategic direction, not a specific source.

0a.1. Start with big problems

Load the audience reference (business-aos/reference/core/audience.md or the relevant offer brief). List 3-5 big problems the target audience has — the chronic, expensive problems that drive them to seek help.

| # | Big Problem |
|---|-------------|
| 1 | [Problem] |
| 2 | [Problem] |
| 3 | [Problem] |

These are trees. MicroMagnets don't solve trees. But every tree has splinters.

0a.2. Break each problem into splinters

For each big problem, ask: "What's one small, annoying piece of this that the person keeps meaning to fix and never does?"

Each big problem should produce 2-3 splinters. A splinter is:

One sentence to describe
Been sitting undone for weeks or months
Would take 1-3 hours to do without AI
Standalone — doesn't require solving something else first
Core action solvable in under 2 minutes once the skill is installed (not "30 minutes of sorting," not "a weekend of work")

| Big Problem | Splinter 1 | Splinter 2 | Splinter 3 |
|-------------|-----------|-----------|-----------|
| [Problem 1] | [Splinter] | [Splinter] | [Splinter] |
| [Problem 2] | [Splinter] | [Splinter] | [Splinter] |
| [Problem 3] | [Splinter] | [Splinter] | [Splinter] |

0a.3. Filter: Output, not diagnosis

This is the most important filter. For each splinter, ask: "What does the person HOLD IN THEIR HANDS after this is solved?"

Kill any candidate where the answer is:

A score
An assessment
An audit
A diagnosis
A report card
A readiness check
A gap analysis
An "understanding" of something

MicroMagnets produce THINGS people USE. The only acceptable outputs are:

Delivery Vehicle	Example
Template	A filled-in LinkedIn About section, a completed email sequence
Script	A word-for-word conversation script, a call opener
Swipe file	A set of ready-to-use headlines, subject lines, DM openers
Cheatsheet	A one-page reference they pin to their wall or bookmark
Calculator	A filled-in model that gives them a number they act on

If a splinter's natural solution is "diagnose the problem" or "assess where you stand" — that's not a MicroMagnet. Kill it. Move on.

Test question: "Could someone copy-paste the output and use it in the next 10 minutes?" If the answer isn't yes, it's not a MicroMagnet.

0a.4. Title heat check

For every surviving candidate, produce a working title. The title must hit all three:

A specific numeric or structural marker ("60-Second," "3-Question," "15-Minute," "4-Filter")
A strong action verb (Kills, Ends, Saves, Reveals, Turns, Catches, Stops)
A specific pain or outcome named explicitly (not "better results")

Passes: "The 4-Question Wrong-Fit Filter That Holds Your No" / "The 15-Minute Tier Map That Ends the Over-Promise Dilemma" / "The 3-Line Snapshot That Kills the Catch-Me-Up Call"

Fails: "Meeting Distiller" (no numeric marker, no specific outcome) / "Scope Anchor" (no marker, no verb, no outcome) / "The Better Review Process" (no marker, weak verb, vague outcome)

If a title can't hit all three, rewrite or kill the candidate.

0a.5. Output candidate list

## MicroMagnet Candidates

| # | Working Title | Splinter | Output (the thing) | Delivery Vehicle | Pre-AI Precedent |
|---|---------------|----------|--------------------|-----------------|-------------------|
| 1 | [Title] | [Splinter] | [What they hold] | Template / Script / etc. | [What people paid for] |
| 2 | [Title] | [Splinter] | [What they hold] | Template / Script / etc. | [What people paid for] |

Kathryn picks. The selected candidate then runs through the Validation Gates below.

Step 0b: Source-driven intake

Use when starting from specific source material — client transcript, voice memo, op-ed, someone else's training.

0b.1. Classify the source mode

Mode	Signal	Special rule
Client conversation	First-person practitioner dialogue, specific engagement detail	Confidentiality filter runs first. Scrub names, financial figures, identifying details. Zero direct quotes in any output. Paraphrase only.
Own thinking	Kathryn's first-person voice, unstructured, exploratory	Extract directly — she's the source of truth.
Op-ed / counterpunch	Third-party author making claims Kathryn would dispute	Extraction generates her counter-POV — what's true that the author missed, what they got wrong, where her audience's reality differs.
Undistilled content	Educational or instructional material from another expert	Default to first-pass inference. Ask for guided extraction only if the source is ambiguous enough that first-pass would be low-confidence.

0b.2. Run the Factory pattern on the source

Extract up to 3 problems × 3 splinters (up to 9 candidates). Not every source fills 9 slots. Empty cells get flagged, not fabricated.

For client conversation sources: Run the audience-stage translation check on every candidate. If the source is from a practitioner above the target audience (a $1M+ firm, an established team, a mature practice), the splinters extracted are real for that source but often mis-aimed for the intended audience. Every splinter must be translated to what the target audience — sub-$500K solo professional services practitioner — would feel today. If the splinter doesn't translate, kill it.

0b.3. Apply promotion criteria

Promote to the candidate list only splinters that pass all five:

Source support — specific evidence in the source, not inferred
Splinter size — core action solvable in under 2 minutes once the skill is installed
Title heat — passes all three tests in 0a.4 above
Operational — friction, not desperation. Practice Builders audience feels bandwidth and memory pressure, not "no leads" terror.
CTA destination exists — the gap points somewhere the Config tilt or menu can address.

Promote the top N that pass all five. N is up to 5, minimum 3. If fewer than 3 pass, stop and tell Kathryn: "The source supports fewer than 3 micromagnet candidates. Consider a different source or a different extraction angle."

0b.4. Output candidate list

Same format as 0a.5. Kathryn picks. Selected candidate runs through the Validation Gates.

Validation Gates

Complete these before building. If any gate fails, stop and pick a different micro-problem or redesign.

Gate 1: Splinter Test

Question	Must Answer
Can you describe the problem in one sentence?	Yes — if it takes a paragraph, it's too big
Is this something they've been meaning to do for weeks/months?	Yes — persistent, not urgent
Would solving it take them 1-3 hours without AI?	Yes — not 5 minutes (too small) and not 2 days (too big)
Is it a standalone problem — not a step in a bigger process?	Yes — solving it doesn't require solving something else first
Is the core action inside the skill solvable in under 2 minutes?	Yes — the skill interaction is a cheat-code moment, not a planning session

Gate 2: Output Test

Question	Must Answer
After running the skill, does the user have something they use immediately?	Yes — paste, send, post, or apply today
Is the output a THING (text, document, template, script) — not a SCORE or ASSESSMENT?	Yes — MicroMagnets produce, they don't diagnose
Could someone look at the output and say "I would have paid for this"?	Yes — per the operating law quality bar: "feel fortunate they got it free"

Gate 3: Pre-AI Precedent

Question	Must Answer
Before AI, did people hire someone or buy a product that did this?	Yes — name the precedent and approximate price
Is the AI version as good or better than the paid version?	Yes — if it's a downgrade, the perceived value drops

Gate 4: Cannibalization Check

Question	Must Answer
Does the paid offer (Practice Builders, Sprint, AOS) sell something that solves this same problem?	No — if yes, you're giving away the store
If someone uses this MicroMagnet, do they still need Practice Builders?	Yes — the MicroMagnet solves a splinter. Practice Builders builds the infrastructure. Different problems.
Could someone chain 5 MicroMagnets together and skip the paid offer?	No — each MicroMagnet is a standalone win. The paid offer solves a fundamentally different class of problem.

Gate 5: Zero-Friction Input

Question	Must Answer
Does the user already have the input data?	Yes — no assembling, no exports, no homework
Can they paste it in under 2 minutes?	Yes — copy from their head, their inbox, or their website
Does it work with messy, unpolished input?	Yes — the whole point is they don't need to prep
Is there a minimum input threshold?	Define it — what's the least they can provide and still get useful output?

Gate 6: Completeness

Question	Must Answer
After running the skill, is the user done?	Yes — no "now do step 2"
Do they need anything else to use the output?	No — the output is self-contained and immediately usable

Gate 7: The Gap (Over-Delivery Gate)

Question	Must Answer
Can you name one adjacent problem this deliberately leaves unsolved?	Yes — the Gap must be explicit, not vague
Does that unsolved problem match what a CTA destination addresses?	Yes — the Gap is what the next step logically solves
Would a prospect who used this and felt it work recognize the Gap on their own?	Yes — the Gap should feel obvious in retrospect, not invented to justify a sales pitch

If you can't name a clear Gap, the candidate fails. Kill it or redesign the splinter to be narrower. The Gap is mandatory — it's the over-delivery gate that keeps the CTA logical.

Positioning (Niche vs. Universal vs. Both)

After validation passes, decide positioning before the Design Brief.

Niche version: the splinter as the source surfaced it, named for the specific practice type. Reserve for one-to-one client work, narrow paid traffic, or vertical-specific LinkedIn campaigns.

Universal version: the same splinter translated up one layer of abstraction to apply to any sub-$500K professional services practitioner. Reserve for top-of-funnel, broad LinkedIn, or cold traffic.

Both: produce both versions in the same build. Doubled outputs. Use when the asset will run on both vertical-specific and broad audiences.

N/A rule. If the splinter doesn't generalize — if it's inherently bound to one vertical's specific processes — the Universal version is "N/A — splinter is inherently niche-bound." Don't force a translation. Ship as niche-only. State the reason in the Design Brief.

Universality test: would a solo consultant, fractional operator, boutique agency owner, or therapist recognize this pattern in their own practice, even if their specifics differ? If yes, Universal is viable. If the pattern requires specific industry knowledge or processes to even understand, it's niche-bound.

Loading Before You Build

Before writing ANY part of the skill file, load:

business-aos/reference/core/voice.md — tone, cadence, vocabulary
business-aos/reference/core/operating-law.md — Win Stack, quality bars, design constraints
business-aos/reference/core/audience.md — or the relevant offer's audience brief
The golden example skill file: content/business/marketing/campaigns/practice-builders-os/wip/linkedin-about-rewriter-skill.md
The voice anchor files from Config
02-micromagnet-terminology.md — locked terms

The Design Brief

After passing all gates and deciding positioning, capture the design in this format before building:

## MicroMagnet Design Brief

**Name:** [Working name]
**Micro-problem:** [One sentence]
**Pre-AI precedent:** [What people paid for + approximate price]
**Who it's for:** [Specific audience — not "everyone"]

**Positioning:** Niche / Universal / Both
  - **Niche frame:** [Practice-type-specific, or "N/A — this candidate is Universal only"]
  - **Universal frame:** [Any-sub-$500K-practitioner, or "N/A — splinter is inherently niche-bound. Reason: [...]"]

**Title:** [Using Taki formula — must pass title heat check]
**Framework name:** [The ™ framework — consistent across Niche and Universal]

**Input:** [What they paste — be specific]
**Output:** [What they get back — the thing they use]

**Core principle:** [The one rule — what's the one thing that, if the skill skips it, the output is garbage?]

**3 Jobs:**
1. [Job name] — [What this part of the output does for them]
2. [Job name] — [What this part does]
3. [Job name] — [What this part does]

**Output sections:**
| # | Section | What It Contains |
|---|---------|-----------------|
| 1 | [Section] | [Description] |
| 2 | [Section] | [Description] |
| ... | ... | ... |

**The Gap (adjacent problem left unsolved):** [Explicit — name the one problem this deliberately doesn't solve, and which CTA destination it points to]

**CTA Destination:** [Low-ticket / $27 workshop / $97mo recurring / Sibling hand-raiser / Future monthly Build]

**Nano spec:** [Which step of the cheat code gets extracted standalone, and what the standalone asset is — template, one-page reference, script snippet]

**Cannibalization check:** [Why this doesn't give away the store]
**Win Stack hits:** [Which of the 5 wins does this hit?]

**Source mode (if applicable):** [Client conversation / Own thinking / Op-ed / Undistilled content — affects confidentiality and voice rules downstream]