Content Capture - From Work to Published Content
Purpose
You just built something, solved a problem, or discovered an insight. It's interesting, but you don't know how to talk about it yet. This skill acts as a journalist interviewing you to extract:
- •What you accomplished and why it matters
- •Who else cares and how to frame it
- •What visuals prove it worked
- •Draft content ready for publishing
Problem solved: "I do interesting work but it never becomes content" Time: 10-18 minutes from session work to ready-to-publish draft
When to Use
Use /capture-content when you've:
- •✅ Solved a significant problem (not trivial bug fixes)
- •✅ Built or updated a system
- •✅ Discovered an interesting tool/workflow through hands-on testing
- •✅ Had a meaningful realization with concrete examples
- •✅ Completed interesting analysis or synthesis work
- •✅ Work is 20+ minutes of substantive effort
- •✅ Content can be shared publicly
When NOT to Use
Skip /capture-content when:
- •❌ Session was purely exploratory (no concrete outcome)
- •❌ Work is confidential, under NDA, or company-proprietary
- •❌ You haven't actually completed anything yet (premature capture)
- •❌ Less than 10 minutes of substantive work
- •❌ Content would require extensive private context to understand
- •❌ Just reading documentation (no original work produced)
- •❌ Creating long-form blog posts (use direct writing instead)
- •❌ Quick status updates (write directly, no interview needed)
Quick Reference
| Scenario | Use? | Reason |
|---|---|---|
| Built feature in 2+ hours | ✅ | Substantial work, clear outcome |
| Fixed interesting bug | ✅ | Problem-solution story |
| Evaluated 3+ tools hands-on | ✅ | Comparative insights |
| Read documentation | ❌ | No original work |
| NDA/proprietary project | ❌ | Cannot share publicly |
| 10-min quick fix | ⚠️ | Only if broadly applicable |
| Tool discovery without testing | ❌ | Insufficient depth |
| Session still in progress | ❌ | Wait for completion |
Interview Process
The skill asks 8-12 questions across 4 phases:
Phase 1: What Happened (3-4 questions)
- •What were you working on?
- •What specific tools/approaches did you use?
- •Walk me through the key moment
Phase 2: Why It Matters (2-3 questions)
- •What surprised you?
- •Why is this different from usual?
- •Who else struggles with this?
Phase 3: Content Strategy (2-3 questions)
- •What's the single most important insight?
- •Is this "behind-the-scenes" or "framework teaching"?
- •What would make someone else care?
Phase 4: Visual Capture (1-2 questions)
- •What should be screenshotted?
- •Would a screen recording add value?
See reference.md for complete interview flow example.
Output Formats
After the interview, generates:
Twitter/X Thread (7-12 tweets)
- •Hook, process breakdown, key insight, call to action
- •Character validation shows over-limit tweets
- •Publishing via bird CLI integration
LinkedIn Post (1,800-2,100 characters)
- •Concrete opening, process, insight, implication
Quote-Tweet Addition (for synthesis content)
- •Technical commentary, implementation experience
Screenshot/Video Script
- •What to capture, how to annotate, what to highlight
Usage
/capture-content # Full interview /capture-content quick # Shorter version (6-8 questions) /capture-content synthesis # For synthesis/framework content /capture-content "solved bug" # With context hint
Publishing Workflow
- •Interview (3-5 min): Answer questions in Claude Code
- •Draft generation: Skill generates content + character analysis
- •Validation:
bird-save-draft.shshows over-limit tweets - •Review: Trim over-limit tweets, capture visuals
- •Publish:
bird-publish-draft.shposts threaded tweets - •Track (24-48h later):
bird-track-engagement.shmeasures performance - •Optimize: High performers feed into synthesis priority queue
See examples.md for complete workflow examples.
Common Mistakes
1. Capturing too early
- •❌ Running
/capture-contentbefore work is complete - •✅ Wait until you have a concrete outcome or insight
2. Over-explaining private context
- •❌ "Our proprietary system for handling X that does Y..."
- •✅ "A caching system that solved Z problem..."
3. Generating content without clear audience
- •❌ "I did this thing and it was interesting"
- •✅ Interview guides you to "Who else struggles with this?"
4. Skipping character validation
- •❌ Publishing tweets over 280 characters (they get rejected)
- •✅ Always run
bird-save-draft.shfirst to see character counts
5. Not tracking engagement
- •❌ Publishing once and never reviewing what worked
- •✅ Track after 24-48h to inform next topics
6. Answering generically
- •❌ "I used AI tools to help" (vague, no insight)
- •✅ "I used Claude Code + bird CLI + jq for..." (specific, actionable)
7. Skipping visuals
- •❌ Text-only threads without proof
- •✅ Screenshots show the actual work, not stock images
Integration
Automatic Detection:
- •Hook detects 3+ signals in session transcript ("solved", "created", "implemented")
- •Prompts at session end: "🎯 CONTENT CAPTURE OPPORTUNITY DETECTED"
- •Suggests running
/capture-contentwhile context is fresh
Publishing Tools:
- •
bird-save-draft.sh- Character validation for Twitter limits - •
bird-publish-draft.sh- Threaded tweet publishing via bird CLI - •
bird-track-engagement.sh- Metrics tracking and synthesis priorities
Performance Optimization:
- •High engagement (50+ score) → Added to synthesis priority queue
- •Signals validated audience interest
- •Prioritize similar topics for next content
Requirements
- •Claude Code v2.1.0+
- •bird CLI v0.5.1+ (
npm install -g @steipete/bird) - •jq (
brew install jq) - •Twitter/X account (logged in via browser for cookie auth)
Installation
See ../README.md for complete installation instructions.
Troubleshooting
See troubleshooting.md for common issues and solutions.
Credits: Based on elrolio's content capture system License: MIT Category: workflows/content-capture