227 lines
9.0 KiB
Markdown
227 lines
9.0 KiB
Markdown
---
|
|
description: Guide you through setting up AI conversation exports for skills analysis
|
|
---
|
|
|
|
# Skills Analysis Setup Guide
|
|
|
|
I'll walk you through setting up your AI conversation exports for analysis and help you get the most value from the Claude Skills Analyzer.
|
|
|
|
## 📋 What You'll Need
|
|
|
|
To analyze your AI usage patterns and generate Custom Skills, you need:
|
|
|
|
1. **AI conversation exports** (Claude and/or ChatGPT)
|
|
2. **Organized directory structure** for your data
|
|
3. **At least 20+ conversations** for basic analysis (50+ recommended for robust patterns)
|
|
|
|
## 📥 How to Export Your Conversations
|
|
|
|
### For Claude Conversations:
|
|
|
|
**Step 1: Request Export**
|
|
1. Go to [claude.ai/settings](https://claude.ai/settings)
|
|
2. Click on **"Privacy"** in the left sidebar
|
|
3. Look for **"Request data export"** or similar option
|
|
4. Click the button to request your export
|
|
|
|
**Step 2: Wait for Email**
|
|
- Claude will send you an email when your export is ready
|
|
- This typically takes 24 hours (sometimes sooner)
|
|
- The email will contain a download link
|
|
|
|
**Step 3: Extract and Save**
|
|
1. Download the ZIP file from the email
|
|
2. **Option A - Automatic** (Recommended): Place the ZIP file in your project directory and run `/extract-exports` - it will automatically extract and organize everything for you
|
|
3. **Option B - Manual**: Extract all files (you'll have `conversations.json`, `projects.json`, etc.) and save them to your `data-exports/claude/` folder (we'll create this next)
|
|
|
|
**Files you should see:**
|
|
- `conversations.json` (required) - Your conversation history
|
|
- `projects.json` (optional) - Project information and metadata
|
|
- `users.json` (optional) - Account information
|
|
|
|
**⚠️ Important Note:**
|
|
Claude exports from claude.ai **DO NOT include Claude Code conversations**. Only conversations from the web interface (claude.ai) are included in the export. If you primarily use Claude Code, your export may have fewer conversations than expected.
|
|
|
|
### For ChatGPT Conversations:
|
|
|
|
**Step 1: Request Export**
|
|
1. Go to [chatgpt.com/settings/general](https://chatgpt.com/settings/general)
|
|
2. Scroll down to **"Data controls"** section
|
|
3. Click **"Export data"**
|
|
4. Select what you want to export (conversations are usually pre-selected)
|
|
5. Confirm the export request
|
|
|
|
**Step 2: Wait for Email**
|
|
- ChatGPT will send you an email when your export is ready
|
|
- This usually takes 2-4 hours
|
|
- The email will contain a download link
|
|
|
|
**Step 3: Extract and Save**
|
|
1. Download the ZIP file from the email
|
|
2. **Option A - Automatic** (Recommended): Place the ZIP file in your project directory and run `/extract-exports` - it will automatically extract and organize everything for you
|
|
3. **Option B - Manual**: Extract all files (you'll have `conversations.json`, `user.json`, etc.) and save them to your `data-exports/chatgpt/` folder (we'll create this next)
|
|
|
|
**Files you should see:**
|
|
- `conversations.json` (required) - Your conversation history
|
|
- `user.json` (optional) - Account information
|
|
- `shared_conversations.json` (optional) - Shared conversations metadata
|
|
- `message_feedback.json` (optional) - Your feedback on responses
|
|
|
|
### Troubleshooting Export:
|
|
|
|
**Can't find Settings?**
|
|
- Claude: Look for your account icon (bottom left) → Settings → Privacy & Data
|
|
- ChatGPT: Click your account name (bottom left) → Settings → scroll to Data controls
|
|
|
|
**Export not arriving?**
|
|
- Check spam/junk email folder
|
|
- Wait a bit longer (can take up to 24 hours)
|
|
- Try requesting again if it's been >24 hours
|
|
|
|
**Files look different?**
|
|
- Export formats can vary slightly between updates
|
|
- As long as you have `conversations.json`, the analysis will work
|
|
- Missing optional files is fine
|
|
|
|
---
|
|
|
|
## 🗂️ Directory Structure Setup
|
|
|
|
I can create the proper directory structure for you automatically. Here's what will be created:
|
|
|
|
```
|
|
your-project/
|
|
├── data-exports/
|
|
│ ├── claude/ # Place Claude exports here
|
|
│ └── chatgpt/ # Place ChatGPT exports here
|
|
├── reports/ # Analysis reports will appear here
|
|
└── generated-skills/ # Generated skills will appear here
|
|
```
|
|
|
|
**Ready to create these directories?**
|
|
|
|
Just confirm and I'll set up:
|
|
- ✅ `data-exports/claude/` - for your Claude conversation exports
|
|
- ✅ `data-exports/chatgpt/` - for your ChatGPT conversation exports
|
|
- ✅ `reports/` - for timestamped analysis reports
|
|
- ✅ `generated-skills/` - for your generated Custom Skills
|
|
|
|
Confirm when you're ready!
|
|
|
|
**💡 Quick Tip**: Once you have the ZIP files from Claude/ChatGPT, you can use the `/extract-exports` command to automatically extract and organize everything for you! Just place the ZIP files in your project directory and run the command.
|
|
|
|
## ⚡ Quick Start Checklist
|
|
|
|
**Setup Phase:**
|
|
- [ ] Directories created (`data-exports/`, `reports/`, `generated-skills/`)
|
|
- [ ] At least one platform export requested (Claude and/or ChatGPT)
|
|
- [ ] Waiting for export email to arrive (24 hours or less)
|
|
|
|
**Data Phase:**
|
|
- [ ] Export files downloaded and extracted
|
|
- [ ] JSON files placed in correct `data-exports/` subdirectories
|
|
- [ ] You have at least 20+ conversations (more is better for patterns)
|
|
|
|
**Analysis Phase:**
|
|
- [ ] Ready to run `/analyze-skills` when exports are ready
|
|
- [ ] Know which output option you want (A, B, C, or D)
|
|
- [ ] Have time for analysis to complete (2-5 minutes typically)
|
|
|
|
**Completion:**
|
|
- [ ] Analysis finished and reports generated
|
|
- [ ] Review skills recommendations
|
|
- [ ] Generate or implement skills as desired
|
|
|
|
## 🎯 What the Analysis Will Find
|
|
|
|
The plugin will identify patterns like:
|
|
|
|
### Business & Communication
|
|
- Email drafting templates
|
|
- Proposal writing workflows
|
|
- Client communication patterns
|
|
- Meeting preparation structures
|
|
|
|
### Development & Code
|
|
- Code review methodologies
|
|
- Documentation standards
|
|
- Debugging approaches
|
|
- Architecture decision patterns
|
|
|
|
### Content & Writing
|
|
- Blog post structures
|
|
- Newsletter formats
|
|
- Social media workflows
|
|
- Research methodologies
|
|
|
|
### Personal Productivity
|
|
- Task planning approaches
|
|
- Decision-making frameworks
|
|
- Learning note-taking systems
|
|
- Goal-setting patterns
|
|
|
|
## 📊 Analysis Options
|
|
|
|
When you run `/analyze-skills`, you'll choose from:
|
|
|
|
- **Option A**: Analysis report only (insights and recommendations)
|
|
- **Option B**: Complete implementation package (ready-to-use skills)
|
|
- **Option C**: Incremental implementation (top 3-5 skills)
|
|
- **Option D**: Custom specification (your defined requirements)
|
|
|
|
## 🔒 Privacy & Security
|
|
|
|
- **Local processing**: All analysis happens on your machine
|
|
- **No data upload**: Your conversations never leave your system
|
|
- **Anonymized output**: Generated skills remove personal information
|
|
- **Git protection**: Export files are automatically ignored by version control
|
|
|
|
## 🚀 Expected Results
|
|
|
|
After analysis, you'll get:
|
|
|
|
### Analysis Reports (`reports/timestamp/`)
|
|
- **Comprehensive analysis** with pattern evidence
|
|
- **Implementation guide** with deployment roadmap
|
|
- **Processing log** for incremental future runs
|
|
|
|
### Generated Skills (`generated-skills/skill-name/`)
|
|
- **SKILL.md** - Main skill with YAML frontmatter
|
|
- **reference.md** - Detailed methodology
|
|
- **examples.md** - Usage examples
|
|
- **templates/** - Reusable output templates
|
|
|
|
## ❓ Common Questions
|
|
|
|
**Q: How many conversations do I need?**
|
|
A: Minimum 20-30 for basic analysis, 50+ for meaningful patterns, 100+ for comprehensive insights.
|
|
|
|
**Q: Are Claude Code conversations included in exports?**
|
|
A: No. Claude exports from claude.ai only include web interface conversations. Claude Code conversations are stored separately and are not included in data exports. If you primarily use Claude Code, consider using ChatGPT exports or the web-based workflow-pattern-analyzer skill instead.
|
|
|
|
**Q: Can I analyze both Claude and ChatGPT together?**
|
|
A: Yes! The plugin performs smart cross-platform deduplication and creates unified skills.
|
|
|
|
**Q: What if I don't have many conversations?**
|
|
A: Start with what you have. The plugin supports incremental processing, so you can re-run analysis as you accumulate more conversations.
|
|
|
|
**Q: How long does analysis take?**
|
|
A: Typically 2-5 minutes for 100 conversations, longer for larger datasets or complex patterns.
|
|
|
|
## 🛠️ Next Steps
|
|
|
|
1. **Set up your directories**: Create the folder structure above (or let `/extract-exports` do it automatically)
|
|
2. **Export your data**: Follow the export guides for your platforms
|
|
3. **Organize your exports**: Run `/extract-exports` to automatically extract and organize ZIP files, or manually place JSON files in the appropriate directories
|
|
4. **Run the analysis**: Use `/analyze-skills` when ready
|
|
5. **Implement skills**: Start with the highest-impact recommendations
|
|
|
|
## 💡 Pro Tips
|
|
|
|
- **Export regularly**: Update your analysis monthly as you accumulate more conversations
|
|
- **Start small**: Begin with Option C (incremental) to test with your top skills
|
|
- **Customize skills**: Edit generated skills to match your specific needs
|
|
- **Share patterns**: Generated skills are great for team standardization
|
|
|
|
Ready to get started? Let me know if you'd like help creating the directory structure or have questions about the export process!
|