Content blitz

Content Blitz · Atomic Content Workflow

Why This Skill Exists

The Operating Framework's content-engine was designed for the 4.6 era—requiring multi-turn conversations (write prompt → generate → review → refine). With Claude 4.7 + 1M context, the entire flow can complete in a single session. This is especially friendly for low-conscientiousness founders: no weekly maintenance rhythm needed, batch generate 30 days of inventory in one shot.

When to Use

• A project just delivered, need to feed back into the acquisition layer
• Need to batch produce a week/month of content
• Have an inspiration burst and want to "dump everything in your head into content at once"

Do not use: without a brand voice file (build the voice section in docs/system-memory.md first).

Input Checklist (must be complete)

1. Content source (at least one):

   • A delivered project → read products/<project>/ directory
   • A theme / insight / story → founder pastes it in
   • A long-form article → read file path

2. Brand voice: voice snippet from docs/system-memory.md

3. Target platforms (all selected by default, can trim):

   • Xiaohongshu (digital human short video script + copy)
   • WeChat moments
   • WeChat blog (optional)
   • Video account (optional)

4. Target cadence:

   • 15 days / 30 days / 60 days

Output (all produced in one session)

A. Master file: products/<project>/content-pack.md

Contains all versions + publishing calendar.

B. Publishing calendar table

Date	Platform	Title	Content Type	Draft Link
Day 1	Xiaohongshu	...	Digital human 30s	#section 1
Day 2	Moments	...	Text	#section 2
...	...	...	...	...

C. Individual drafts (each = one section)

For each date and piece:

• Title (platform-native)
• Body (brand voice compliant)
• Digital human script (if video): includes shot cues + lines + image suggestions
• Hashtags (Xiaohongshu: 5-8 required)
• CTA: direct to WeChat / WeChat blog

Content Structure Principles

70-20-10 Mix

For 30 days = 30 pieces:

• 70% technical value (21 pieces): niche SaaS tools, AI applications, specific operations
• 20% case stories (6 pieces): anonymized delivery stories, measurable results
• 10% personal (3 pieces): city life, learning notes, tool comparisons

Viral Templates (for Xiaohongshu)

Generate using these structures by default, cycle through them:

• Counterintuitive hook + numerical evidence: e.g., "The dashboard I spent 3 hours on—the seller said after 1 week they no longer needed an ad team"
• VS comparison: traditional method vs AI method, side-by-side
• Before/After: specific before/after changes + data
• Tutorial: step-by-step with demo GIF / screenshots
• Pitfall guide: N traps I fell into

Digital Human Script Format

[Shot 1 · 0-5 sec] Person center frame + hook subtitle
Lines: "Today I'll tell you..."
Image suggestion: <specific>

[Shot 2 · 5-15 sec] Problem scenario
Lines: "If you are..."
Image suggestion: <specific>

[Shot 3 · 15-25 sec] Solution demonstration
Lines: "Here's what I did..."
Image suggestion: <demo screenshot / data>

[Shot 4 · 25-30 sec] CTA
Lines: "DM me on my homepage to chat"
Image suggestion: <WeChat QR code / homepage guide>

Quality Checklist (run after generation)

• All content passes brand voice check (no "empower the ecosystem" fluff)
• At least 1 specific number (time, amount, percentage) every 5 pieces
• Every CTA points to the same destination (WeChat / DM)
• No real customer names or unauthorized data leaked
• Xiaohongshu hashtags are all relevant (no unrelated trending tags stacked)
• Digital human script lines read naturally (can be delivered within 30 seconds)

Execution Flow

Input: project path OR theme + days + platform list
   ↓
Read: voice section from docs/system-memory.md
Read: input source (project products directory / user pasted text)
   ↓
One-shot generate: publishing calendar + all drafts (all 30 pieces)
   ↓
Self-check: run quality checklist
   ↓
Write: products/<project>/content-pack.md
   ↓
Output: founder receives full content pack + one recommendation (which to post first, why)

Limitations / When Not to Use

• Not suitable for real-time data content (trend tracking, holiday tie-ins)—that needs a separate run
• Not suitable for serious knowledge products (course scripts)—precision requirements too high
• Not suitable for companies with existing content teams—multi-person review workflow conflicts with this

Cost Note

One run of 30 pieces: Opus approximately ¥20-40 (includes 1M context loading and output). Batch production vs piecemeal 3 pieces at a time: cost is similar, but your mental load is 10x lower. This is the best value investment in the 4.7 era.

Prohibitions

• Do not run without a brand voice file (refuse execution directly, require building it first)
• Do not automate publishing (generate drafts only, you handle publishing)
• Do not mix content across projects (case A's story is not reused for project B)