MicroMagnet Factory — Production Workflow

How to Use This Kit

Follow this workflow in order. Do not skip the validation gates. Do not build the skill file before the design brief is complete.

Step 0: Intake (If Needed)

If Kathryn names the micro-problem directly, skip to Step 1.

Otherwise, pick the intake path that matches the source.

Step 0a: Audience-theory intake

Use when starting from strategic direction without specific source material.

Run the full generation process from 01-micromagnet-context.md → Step 0a. Lists big problems from the audience reference, breaks each into splinters, filters for output (not diagnosis), runs the title heat check, and presents candidates.

Kathryn selects the candidate. Nothing moves to Step 1 without her pick.

Step 0b: Source-driven intake

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

Run the full process from 01-micromagnet-context.md → Step 0b. Classifies source mode (and runs confidentiality filter first for client conversations), extracts up to 9 candidates via the Factory pattern, applies the 5 promotion criteria (source support, splinter size, title heat, operational, CTA destination exists), and presents the candidates that pass.

For client-conversation sources, the audience-stage translation check runs on every candidate — splinters that don't translate to the target audience get killed, not stretched.

Kathryn selects the candidate. Nothing moves to Step 1 without her pick.

Step 1: Capture the Micro-Problem

Whether Kathryn named the problem directly or selected it from generated candidates, capture it before proceeding:

• The problem in one sentence
• Who has it (specific audience)
• How long it typically goes unsolved
• What people do instead (workaround, hire someone, ignore it)
• What the output IS — the specific thing they hold in their hands after running

Do NOT proceed if:

• The problem takes a paragraph to explain (too big)
• The problem is "I need a system for X" (that's infrastructure, not a splinter)
• The solution is "understand X better" (that's education, not output)
• The solution is "find out where X is broken" (that's a diagnostic, not a MicroMagnet)
• The solution is a score, assessment, audit, or readiness check (those diagnose — MicroMagnets produce)
• The core action takes more than 2 minutes once the skill is installed (that's a planning tool, not a MicroMagnet)

Step 2: Validate Against MicroMagnet Criteria

Run all seven gates from 01-micromagnet-context.md:

1. Splinter test — one tiny persistent problem, core action under 2 minutes
2. Output test — produces a thing, not an analysis
3. Pre-AI precedent — people used to pay for this
4. Cannibalization check — doesn't overlap with paid offers
5. Zero-friction input — they already have it, paste in under 2 minutes
6. Completeness — they're done after running
7. The Gap — one adjacent problem named explicitly, pointing at a specific CTA destination

If any gate fails, stop. Pick a different micro-problem or redesign. Do not build a skill that fails a gate — it will produce something nobody uses, something that undercuts the business, or something with a CTA that makes no sense.

Step 3: Decide Positioning and Write the Design Brief

Decide positioning (Niche / Universal / Both) using the test in 01-micromagnet-context.md. Universal requires that a non-niche practitioner would recognize the pattern in their own practice. If the splinter is inherently niche-bound, mark Universal as N/A and state the reason.

Complete the design brief template from 01-micromagnet-context.md. This is your build spec.

Load before writing:

• business-aos/reference/core/voice.md
• business-aos/reference/core/operating-law.md
• The relevant audience reference (AOS audience or PB audience)
• The voice anchor files named in Config
• The golden example: 03-micromagnet-golden-example.md → then the actual skill file it points to

The design brief must include:

• Name, micro-problem, pre-AI precedent, target audience
• Positioning decision (Niche / Universal / Both) with N/A reason if Universal not viable
• Title (must pass title heat check) and framework name
• Input (what they paste) and output (what they get back)
• Core principle (the one rule — stays in the Kit, not in the shipped skill)
• 3 jobs
• Output sections table
• The Gap — adjacent problem left unsolved, and which CTA destination it points to
• CTA destination
• Nano spec — which step of the cheat code extracts standalone
• Cannibalization check
• Win Stack hits
• Source mode if applicable (affects confidentiality and voice rules)

Present the design brief to Kathryn before building. She validates the concept. Nothing gets built until she says go.

Step 4: Build the Skill File (Prospect-Facing Shipped Output)

Produces: [skill-name]-skill.md

The shipped skill file is a prospect-facing product, not an internal tool. It strips internal scaffolding (Core Principle, standalone Rules, embedded QC, What Makes This Different). Those stay in the Kit. This file is what the prospect receives.

4a. YAML Frontmatter

---
name: [skill-name]
description: >
  [Multi-line description. What it does, what input it takes,
  what output it produces, when to use it.
  Triggers: "trigger 1", "trigger 2", "trigger 3", or any request
  to [description of the job].]
metadata:
  author: Kathryn Brown, Advisory OS
  version: "1.0.0"
  updated: "[YYYY-MM-DD]"
---

4b. Heading + Positioning

# [Skill Name — including the ™ framework name if the title carries it]

[One sentence — what the user walks away with. Not what the skill does internally. What they hold at the end.]

4c. What You Paste

## What You Paste

[Input spec in prospect-facing language. Accepted formats. Tolerance for thin input. Minimum viable input if applicable.]

4d. What You Get Back

## What You Get Back

1. [Deliverable 1 — specific, concrete]
2. [Deliverable 2]
3. [Deliverable 3]
[Up to 5 total. If the skill produces fewer, write fewer — don't pad.]

4e. The Method

## The Method: The [Framework Name]™

### Step 1: [Name]
**Input:** [What this step takes in]
**Output:** [What this step produces]

[Instructions to Claude — how to run this step. Constraints that would otherwise live in a separate Rules section live inline here: "Never rebalance into Week 4," "If input is under X tokens, ask for more," etc.]

### Step 2: [Name]
[...]

### Step 3: [Name]
[...]

Every step names its inputs and its outputs. Every necessary behavior rule lives inline with the step it applies to. No standalone Rules section.

4f. Output

## Output

[Fenced markdown template showing exactly the output the prospect gets back when they run the skill. Complete, copy-pasteable, matches the design brief's output sections exactly.]

4g. The Next Step

## The Next Step

[CTA PowerUp block — 3 sentences max, ending in a link. Pull offer name and URL from Config.]

CTA PowerUp Formulas

Use the matching formula for the destination:

Low-ticket paid offer:

This [micromagnet name]™ handles [what it solves]. The next layer — [The Gap, named explicitly] — is what [Config: Low-ticket offer name] was built for. [Config: Low-ticket offer URL].

$27 workshop:

You just used [micromagnet name]™. The full toolkit it lives inside — Practice Brain plus five more skills — is what we build together in [Config: $27 workshop name]. [One sentence on what they walk away with]. [Config: $27 workshop URL].

$97/mo recurring:

[Micromagnet name]™ is one tool. [Config: Recurring offer name] is the monthly build cycle where you install a new one every month — tools that compound into a full operating system for your practice. [Config: Recurring offer URL].

Sibling hand-raiser:

If [micromagnet name]™ solved [problem], the next thing to grab is [sibling skill name]. It handles [adjacent problem]. [Link].

Future monthly Build:

This is a preview of the kind of tool we build live inside [Config: Recurring offer name]. Next month's Build: [topic]. [Config: Recurring offer URL].

If any Config value needed is [SET ME], stop and flag before producing. Do not ship with placeholder URLs.

4h. Close

---

© 2026 Kathryn Brown, Advisory OS. Licensed for your own practice. Not for redistribution or resale. Contact: kathryn@creatingyourplan.com

Trademark Rule

Framework name gets ™ on first use and on every heading where it appears. Step names inside the framework do NOT get ™.

Step 4.5: Extract the Nano

Build the nano in the same session as the skill. The brief named which step of the cheat code extracts standalone.

Produces: [skill-name]-nano.md

The nano's vehicle follows the friction test — it's often the same vehicle as the micro (a shorter version), sometimes a different one (the micro is a skill, the nano is a cheat sheet pulled from one step).

Nano rules:

• No CTA. No PowerUp block. No "Next Step" section. Nanos are for re-engagement and thank-you use, not acquisition.
• Smaller than the parent micromagnet. A nano that replicates the micro has failed.
• Works standalone. A prospect can use it without running the full micromagnet first.

Structure mirrors the parent micromagnet's vehicle minus The Next Step section.

Step 5: Test the Skill

Start a new Claude conversation. Upload the skill file. Run it against real data.

Minimum 3 tests:

For each test, check:

• All sections present (or correctly skipped)?
• Format matches the template?
• Voice matches the input?
• Zero AI-rhythm violations?
• Within any character/length limits?
• Edge cases handled?

Fix and re-test until output is consistent across all three tests.

Step 6: Build the Quick-Start Guide

Produces: [skill-name]-quick-start.md

Only build this AFTER the skill file passes testing. The quick-start references the tested trigger phrase, input requirements, and output.

Structure:

# [Skill Name] — Quick Start

You just got a free Claude skill. Here's how to use it.

## What You Need

- A Claude Pro account (or higher) — Skills require a paid plan
- [Input description — what they need to have ready]

## Setup (One Time)

### Step 1: Enable Code Execution
[Exact UI path — click your name, Settings, Capabilities, toggle]

### Step 2: Upload the Skill
[Exact button names — Customize icon, Skills, +, Upload a skill, toggle on]

That's it. The skill is installed.

## Running the Skill

1. Start a new conversation in Claude
2. Type: **"[Trigger phrase]"**
3. Paste [input description]
4. Wait about [time estimate]

[1-2 sentences describing what they get back.]

## Tips

**[Tip from testing.]**
[One sentence.]

**[Tip from testing.]**
[One sentence.]

**[Tip from testing.]**
[One sentence.]

## What to Paste

Any of these work:

- [Input type 1]
- [Input type 2]
- [Input type 3]

[One sentence about what the skill handles.]

---
*Built by Kathryn Brown — Advisory OS*

Rules:

• Tips come from real testing, not generic advice
• "What to Paste" lists specific types, not categories
• Every instruction references exact UI elements
• No marketing language

Step 7: Handle "Both" Positioning (If Applicable)

If the design brief selected Both (Niche + Universal), produce six files total:

1. [skill-name]-niche-skill.md
2. [skill-name]-niche-quick-start.md
3. [skill-name]-niche-nano.md
4. [skill-name]-universal-skill.md
5. [skill-name]-universal-quick-start.md
6. [skill-name]-universal-nano.md

The Niche and Universal versions share the same cheat code — the method the micromagnet delivers. What changes:

• Title (niche-specific vs. universal framing)
• Splinter language (practice-type-specific vs. pattern-level)
• Examples (one vertical's scenarios vs. cross-vertical scenarios)
• Some phrasing throughout

The framework stays identical. Don't rewrite the method for each positioning — only the framing around it. The QC checklist includes a divergence check that enforces this (same framework name, same step count across both).

Nanos differ too. A nano extracted from the niche micromagnet uses niche-specific language; a nano extracted from the universal uses universal framing. Don't share one nano across both.

Step 8: QC Both Assets

Run 04-micromagnet-quality.md against all files produced. Fix all failures before shipping.

Blocking checks:

• All 7 MicroMagnet criteria pass (including The Gap)
• Layer separation is clean — no internal scaffolding in the shipped skill
• Skill file tested on 3+ inputs with consistent output
• Quick-start has exact UI paths
• Voice check passes (matches Config voice anchors)
• Zero AI-rhythm violations in test outputs
• Nano is smaller, has no CTA, works standalone
• For Both positioning: framework name and step count identical across Niche and Universal
• No [SET ME] Config placeholders shipped in the CTA PowerUp

Step 9: Deliver

Final files go to the campaign folder: content/business/marketing/campaigns/[campaign-name]/wip/

Move to root of campaign folder when shipped. Update STATUS.md.

Next in the pipeline: Delivery Page Kit → Handraiser Post Kit → DM Kit