aiwg
Version:
Cognitive architecture for AI-augmented software development with structured memory, ensemble validation, and closed-loop correction. FAIR-aligned artifacts, 84% cost reduction via human-in-the-loop, standards adopted by 100+ organizations.
113 lines (84 loc) • 3.15 kB
Markdown
# Droid Bridge Addon - Claude Code Priming
> Add this section to your project's CLAUDE.md to enable Droid orchestration.
---
## Factory Droid Integration
This project has **Factory Droid** available as a secondary AI agent for automated code operations.
### Understanding the Relationship
**Critical Context**: Claude Code and Factory Droid are **independent AI agents** that share:
- The same filesystem
- The same git repository
- The same terminal environment (but not simultaneously)
They do **NOT** share:
- Conversation context or memory
- Real-time coordination
- Direct communication channels
### When to Use Droid vs Claude Code Agents
| Task Type | Use | Reason |
|-----------|-----|--------|
| Quick batch fixes (linting, formatting) | **Droid** | Fast, non-interactive |
| Code review with context | **Claude Code agents** | Has conversation history |
| Auto-fix on commit | **Droid** (automatic) | Pre-commit hook |
| Complex multi-file refactoring | **Claude Code** | Better context management |
| CI/CD fixes | **Droid** (automatic) | Pipeline integration |
| Architecture planning | **Claude Code agents** | Requires dialogue |
### Invoking Droid from Claude Code
Use the `/invoke-droid` command or `droid-launcher` skill:
```bash
# Via command
/invoke-droid "fix all TypeScript errors"
# Via skill (with monitoring)
/droid-launcher "refactor utils.js to use ES6 modules"
```
### Droid Invocation Patterns
**Simple execution** (fire-and-forget):
```bash
droid exec "task description"
```
**With output capture** (for processing results):
```bash
droid exec "task description" 2>&1 | tee .aiwg/logs/droid/last-run.log
```
**Background execution** (for long tasks):
```bash
nohup droid exec "task" > .aiwg/logs/droid/background.log 2>&1 &
echo $! > .aiwg/logs/droid/pid
```
### Monitoring Droid Operations
Use `/monitor-droid` to check on background operations:
- View running Droid processes
- Read output logs
- Check exit status of last run
### File-Based Coordination Protocol
Since Claude Code and Droid cannot communicate directly, use files for coordination:
**Request file** (Claude Code writes):
```
.aiwg/droid/request.json
{
"task": "description",
"timestamp": "ISO-8601",
"requester": "claude-code",
"context_files": ["file1.js", "file2.js"]
}
```
**Response file** (Droid writes via prompt):
```
.aiwg/droid/response.json
{
"status": "complete|error",
"timestamp": "ISO-8601",
"files_modified": ["file1.js"],
"summary": "What was done"
}
```
### Best Practices
1. **Don't duplicate work**: If Claude Code can do it with agents, don't also invoke Droid
2. **Use Droid for speed**: Quick fixes, formatting, simple refactors
3. **Use Claude Code for depth**: Reviews, planning, complex changes
4. **Check results**: Always verify Droid's output - it operates without conversation context
5. **Log everything**: Use `.aiwg/logs/droid/` for audit trail
### Limitations
- Droid cannot see your conversation with Claude Code
- Droid may make different decisions than Claude Code would
- No real-time feedback during Droid execution
- Interactive Droid mode requires separate terminal
---