imehr avatar

capture-content

Use when you've done interesting work but don't know how to articulate it - journalist-style intervi

提供方 imehr|开源

Content Capture - From Work to Published Content

Transform interesting work into publishable content through a 10-18 minute journalist-style interview.

Problem Solved

"I do interesting work but it never becomes content." This skill solves the friction between doing work and sharing it.

What It Does

  1. Automatic Detection: Hook detects interesting work at session end
  2. Guided Interview: 8-12 questions extract what matters and why
  3. Content Generation: Generates Twitter threads, LinkedIn posts, with character validation
  4. Publishing Integration: Scripts for validation, publishing (bird CLI), and engagement tracking
  5. Performance Optimization: High-engagement content feeds synthesis priorities

Time Investment

  • Interview: 3-5 minutes
  • Visual capture: 2-3 minutes
  • Refinement: 5-10 minutes
  • Total: 10-18 minutes from session work to ready-to-publish draft

Quick Start

# Install (from marketplace)
/plugin install content-capture@imehr-marketplace

# Run installation script
cd ~/.claude/skills/content-capture
./install.sh

# Restart Claude Code

# Use it
/capture-content

Usage

/capture-content           # Full interview (8-12 questions)
/capture-content quick     # Shorter version (6-8 questions)
/capture-content synthesis # For synthesis/framework content

When to Use

✅ Use when you've:

  • Solved a significant problem
  • Built or updated a system
  • Discovered a tool/workflow through hands-on testing
  • Had a meaningful realization with concrete examples
  • 20+ minutes of substantive work
  • Content can be shared publicly

❌ Skip when:

  • Work is confidential/NDA
  • Just reading documentation (no original work)
  • Less than 10 minutes of work
  • Session still in progress

See SKILL.md for complete Quick Reference table.

Publishing Workflow

/capture-content
↓
Answer interview questions (3-5 min)
↓
Draft generated with character analysis
↓
bird-save-draft.sh <name> <content> [screenshot]
↓
Review and trim over-limit tweets
↓
bird-publish-draft.sh <draft-folder>
↓
Track engagement (24-48h later)
↓
bird-track-engagement.sh <draft-folder>
↓
High performers → synthesis priority queue

Requirements

  • Claude Code: v2.1.0+
  • bird CLI: v0.5.1+ (install)
  • jq: brew install jq
  • Twitter/X: Logged in via browser (for cookie auth)

Installation

Via Marketplace (Recommended)

/plugin install content-capture@imehr-marketplace
cd ~/.claude/skills/content-capture
./install.sh

Manual Installation

# Clone or download skill files
git clone https://github.com/imehr/skills
cd skills/workflows/content-capture

# Run installer
./install.sh

# Follow prompts to configure hooks

Components

1. Skill (/capture-content)

  • Journalist-style interview (8-12 questions)
  • Extracts: what, why, who-cares, visual proof
  • Generates Twitter threads, LinkedIn posts

2. Detection Hook

  • Automatically runs at session end
  • Scans for 10 signal patterns
  • Prompts when 3+ signals detected

3. Publishing Scripts

  • bird-save-draft.sh - Character validation
  • bird-publish-draft.sh - Threaded publishing
  • bird-track-engagement.sh - Metrics tracking

4. Performance Loop

  • Track engagement after 24-48h
  • High performers (50+ score) → synthesis queue
  • Optimize future topic selection

File Structure

content-capture/
├── SKILL.md                 # Main skill (<1000 words)
├── TEST_RESULTS.md          # Testing documentation
├── README.md                # This file
├── install.sh               # Installation script
├── scripts/
│   ├── hook.sh             # Session detection
│   ├── save-draft.sh       # Character validation
│   ├── publish-draft.sh    # Publishing
│   └── track-engagement.sh # Metrics
└── docs/
    ├── reference.md        # Complete interview flow
    ├── examples.md         # Workflow examples
    └── troubleshooting.md  # Common issues

Documentation

Testing

Tested with RED-GREEN-REFACTOR methodology:

  • ✅ Complex implementation (2+ hours)
  • ✅ Simple tool evaluation
  • ✅ Documentation reading (skip)
  • ✅ False positive prevention
  • ⚠️ Failed experiments (partially tested)

See TEST_RESULTS.md for complete test scenarios.

Example Output

Input: 2 hours implementing content capture system

Interview extracts:

  • Problem: Skill structure format issues
  • Solution: Directory-based with uppercase SKILL.md
  • Insight: Single-file vs directory formats
  • Audience: Others installing custom skills

Generated content: Twitter thread with:

  • Hook: "Spent 2 hours figuring out why my Claude Code skill wouldn't load..."
  • Process: Directory structure requirements
  • Insight: SKILL.md must be uppercase
  • Proof: Working /capture-content command

Integration

With Automatic Detection

Hook runs at session end, prompts when interesting work detected.

With Publishing

Bird CLI integration for Twitter threading, character validation.

With Performance Tracking

Engagement metrics feed synthesis priority queue for topic optimization.

With Other Skills

  • Works with episodic-memory for recalling past captures
  • Complements writing-clearly-and-concisely for long-form content
  • Feeds brainstorming for content strategy

Troubleshooting

Skill not appearing:

  • Ensure directory structure: ~/.claude/skills/content-capture/SKILL.md
  • SKILL.md must be uppercase
  • Restart Claude Code

Hook not triggering:

  • Check ~/.claude/settings.json has hooks configured
  • Verify hook file is executable: chmod +x ~/.claude/hooks/content-capture-trigger.sh
  • Restart Claude Code

Bird CLI authentication:

  • Log into x.com in Chrome/Safari/Firefox
  • Bird uses browser cookies automatically
  • Run bird whoami to verify

See docs/troubleshooting.md for more.

Contributing

Found a bug or have a suggestion?

  1. Test your scenario
  2. Document expected vs actual behavior
  3. Submit issue to imehr/skills

Credits

License

MIT License - See LICENSE file

Category

workflows/content-capture

Version

1.0.0


Ready to turn your work into content? Install and run /capture-content 🚀