Content Creation

Communicate Track — Step 3 of 4. Turns IMC angles into production-ready content with A/B variants.

Inputs Required

Angle + channel from .agents/mkt/imc-plan.md (or user-provided)

Output

.agents/mkt/content/[slug].md

Quality Gate

Before delivering, verify:

Hook communicable in ≤8 words (if you can't summarize what it says in 8 words, it's too complex)
Hook passes 3-question test: visual? falsifiable? uniquely yours?
CTA follows formula: [action verb] + [what they get] (not "Learn More" or "Click Here")
A/B variant changes exactly ONE element — and hypothesis states what you'd learn if B wins

Chain Position

Previous: mkt-imc | Next: mkt-attribution

Before Starting

Step 0: Product Context

Check for .agents/mkt/product-context.md. If available, read for product details and accuracy.

Required Artifacts

Artifact	Source	If Missing
`imc-plan.md`	mkt-imc	INTERVIEW. Ask: angle, channel, placement, awareness stage.

Optional Artifacts

Artifact	Source	Benefit
`icp-research.md`	mkt-icp-research	VoC language for copy
`product-context.md`	mkt-copywriting	Product details for accuracy

Read .agents/mkt/imc-plan.md if it exists — pull angle, hook type, awareness stage, channel. If it doesn't exist, interview:

"What angle are we creating content for?"
"What channel and placement?"
"What awareness stage is the audience at?"

WebSearch directive: Before writing, verify current platform specs: "[platform] [format] specs [year]" (e.g., "LinkedIn carousel specs 2026"). See references/platform-specs.md for baseline reference.

Step 1: Write the Hook

The hook stops the scroll. Write the exact text.

Hook Type	Pattern	Example
Question	"Have you ever [relatable]?"	"Have you ever spent more time reporting than doing?"
Bold claim	"[Controversial statement]"	"Your meetings problem isn't meetings."
How-to	"How to [outcome] without [sacrifice]"	"How to ship 2x faster without hiring"
Story	"I [action] and [surprise]"	"I cut meetings 80%. Output doubled."
Data	"[Surprising stat]"	"31 hours/month in unproductive meetings"
Contrarian	"Stop [common advice]"	"Stop tracking velocity."

Test the hook: Run it through the 3-question test. If any answer is no, rewrite before continuing.

Step 2: Write the Body

Write the ACTUAL copy — not an outline, not a description.

Video: Timestamped script with text-on-screen, voiceover, and visual direction. Static post: Complete caption, every word. Carousel: Every slide, cover to CTA. Thread: Every post, numbered.

Step 3: Write the CTA

Stage	Pattern	Example
Unaware	Discover	"See why this matters →"
Problem Aware	Explore	"Discover the fix →"
Solution Aware	Compare	"See how we're different →"
Product Aware	Try	"Start free trial →"
Most Aware	Act	"Start now — first month free →"

Include: exact CTA text, destination, tracking parameters.

Step 4: A/B Variant

Create variant B that changes exactly ONE element:

What to Test	When	What You Learn if B Wins
Hook	Default — always test the hook first	Audience responds more to [B's approach] than [A's]
CTA	Hook is strong, conversion is weak	[B's framing] reduces friction better
Format	Engagement is low despite good hook	Audience prefers [B's format] on this platform
Angle framing	Testing different emotional entry points	[B's emotion] resonates more than [A's]

Step 5: Asset Checklist

What the production team needs. Be specific about dimensions, formats, and brand assets.

Artifact Template

---
skill: mkt-content-create
version: 1
date: [today's date]
status: draft
---

# Content: [Title]

**Date:** [today]
**Skill:** mkt-content-create
**Angle:** [from IMC plan]
**Channel:** [platform + placement]
**Stage:** [awareness level]

## Hook

[Exact hook text]

**3Q Test:** Visual: [Y/N] | Falsifiable: [Y/N] | Uniquely ours: [Y/N]

## Body

[Complete copy — every word, slide, or timestamp]

## CTA

- **Text:** [exact CTA copy]
- **Destination:** [URL or action]
- **Tracking:** [UTM parameters]

## A/B Variant

| Element | A (Main) | B (Variant) |
|---------|----------|-------------|
| [Hook/CTA/Format] | [Original] | [One change] |

**Test hypothesis:** If B wins, it means [specific learning about the audience].

## Asset Checklist

- [ ] [Asset with specs]
- [ ] [Asset with specs]
- [ ] [Brand assets needed]

> On re-run: rename existing artifact to `[slug].v[N].md` and create new with incremented version.

Worked Example

# Content: Status Update Waste — LinkedIn Carousel

**Date:** 2026-03-13
**Skill:** mkt-content-create
**Angle:** "Engineering teams lose 12 hrs/week to status updates nobody reads"
**Channel:** LinkedIn carousel, @brand
**Stage:** Problem Aware

## Hook

"Your team loses 12 hours a week to status updates nobody reads."

**3Q Test:** Visual: Y (12 hours, status updates) | Falsifiable: Y (12 is a specific, checkable number) | Uniquely ours: Y (cites our research data)

## Body

- Slide 1: "Your team loses 12 hours a week to status updates nobody reads."
- Slide 2: "Here's where those hours go:" [pie chart visual]
- Slide 3: "4 hrs — writing updates for managers who skim them"
- Slide 4: "3 hrs — standups where 8 people wait to give 30-second updates"
- Slide 5: "3 hrs — switching between tools to piece together who's doing what"
- Slide 6: "2 hrs — 'quick syncs' that could have been a dashboard"
- Slide 7: "The fix isn't better updates. It's fewer updates with better visibility."
- Slide 8: CTA slide

## CTA

- **Text:** "See how teams ship without status theater →"
- **Destination:** /product/visibility
- **Tracking:** utm_source=linkedin&utm_medium=organic&utm_campaign=status-waste&utm_content=carousel-v1

## A/B Variant

| Element | A (Main) | B (Variant) |
|---------|----------|-------------|
| Hook | "Your team loses 12 hrs/week..." (Data) | "I stopped doing standups. Here's what happened." (Story) |

**Test hypothesis:** If B wins, our audience responds more to personal narrative than statistics on LinkedIn.

## Asset Checklist

- [ ] 8 slides, 1080×1350px (LinkedIn carousel format)
- [ ] Slide 2: pie chart showing hour breakdown
- [ ] Brand fonts + colors applied to all slides
- [ ] Logo on slides 1 and 8

References

references/examples.md — Complete worked examples with full copy
references/platform-specs.md — Platform dimensions, limits, native patterns