Initial commit

This commit is contained in:
Zhongwei Li
2025-11-30 09:04:23 +08:00
commit d7ebdd4819
30 changed files with 12517 additions and 0 deletions

196
commands/steering-clean.md Normal file
View File

@@ -0,0 +1,196 @@
---
description: Clean up old archives, logs, and temporary files to free disk space
---
# Steering Context Generator - Clean
Remove old archives and temporary files to free up disk space.
## Quick Start
```bash
/steering-clean
```
This will:
- Archive current context (backup)
- Remove archives older than 7 days
- Clean logs older than 30 days
- Remove cache files
- Delete temporary files
## What Gets Cleaned
| Item | Location | Retention | Impact |
|------|----------|-----------|--------|
| Old archives | `.claude/memory/archives/` | 7 days | Can regenerate |
| Old logs | `.claude/logs/` | 30 days | Lost history |
| Cache files | `.claude/steering/v2.0/cache/` | All | Rebuilt on next use |
| Temp files | `.claude/memory/**/*.tmp` | All | Safe to delete |
## Implementation
```bash
bash scripts/cleanup.sh
```
## Expected Output
```
🧹 Cleaning Steering Context Generator artifacts...
Artifacts to clean:
Archives: 450 MB
Logs: 23 MB
Cache: 12 MB
Continue? (y/N) y
Archiving current context to .claude/memory/archives/backup_20251102_120000...
✓ Archived 9 files
Cleaning old archives (>7 days)...
✓ Removed 3 old archives (380 MB freed)
Cleaning old logs (>30 days)...
✓ Removed 145 log files (18 MB freed)
Cleaning cache...
✓ Cleared cache (12 MB freed)
Cleaning temporary files...
✓ Removed 23 temp files (2 MB freed)
Cleanup complete!
Current usage:
Archives: 70 MB
Logs: 5 MB
Total: 250 MB
Total freed: 412 MB
```
## Safety Features
**Automatic Backup**:
Before cleaning, current context is archived:
```
.claude/memory/archives/backup_YYYYMMDD_HHMMSS/
├── ARCHITECTURE.md
├── AI_CONTEXT.md
├── CODEBASE_GUIDE.md
└── ...
```
**Confirmation Prompt**:
Interactive mode asks for confirmation before deleting.
**Dry Run**:
See what would be cleaned without actually deleting:
```bash
bash scripts/cleanup.sh --dry-run
```
## Cleanup Options
### Aggressive Cleanup
Remove all archives (not recommended):
```bash
# Manual aggressive cleanup
rm -rf .claude/memory/archives/*
rm -rf .claude/logs/*
```
### Selective Cleanup
Clean specific areas only:
**Archives only**:
```bash
find .claude/memory/archives -type d -mtime +7 -exec rm -rf {} +
```
**Logs only**:
```bash
find .claude/logs -type f -mtime +30 -delete
```
**Cache only**:
```bash
rm -rf .claude/steering/v2.0/cache/*
```
### Custom Retention
Edit `scripts/cleanup.sh` to change retention periods:
```bash
# Archives: Change from 7 to 30 days
find .claude/memory/archives -type d -mtime +30 -exec rm -rf {} +
# Logs: Change from 30 to 90 days
find .claude/logs -type f -mtime +90 -delete
```
## When to Clean
**Regular Maintenance**:
- Weekly: For active projects
- Monthly: For stable projects
- Before: Major releases
- When: Low disk space warnings
**Signs You Need Cleaning**:
- ⚠ Archive size > 500 MB
- ⚠ Total .claude/ size > 2 GB
- ⚠ Disk space warnings
- ⚠ Slow operations
## Recovering Cleaned Data
**Recent Backup**:
```bash
# List available backups
ls -lh .claude/memory/archives/backup_*/
# Restore latest backup
LATEST=$(ls -t .claude/memory/archives/backup_* | head -1)
cp -r $LATEST/*.md .claude/steering/
```
**From Git**:
If context files are committed:
```bash
git log -- .claude/steering/
git checkout HEAD~1 -- .claude/steering/
```
**Regenerate**:
If no backups available:
```bash
/steering-generate
```
## Troubleshooting
### "Permission denied" errors
```bash
chmod +x scripts/cleanup.sh
```
### Cleanup doesn't free space
Check hidden or locked files:
```bash
lsof | grep .claude
```
### Accidentally deleted current context
```bash
# Restore from latest backup
LATEST=$(ls -t .claude/memory/archives/backup_* | head -1)
cp -r $LATEST/*.md .claude/steering/
```
---
**Free up space:** Run `/steering-clean` regularly!

132
commands/steering-config.md Normal file
View File

@@ -0,0 +1,132 @@
---
description: View and modify Steering Context Generator configuration settings
---
# Steering Context Generator - Configuration
View and customize system configuration.
## Quick Start
**View config**:
```bash
cat .claude/steering/config.json | jq '.'
```
**Edit config**:
```bash
vim .claude/steering/config.json
```
## Default Configuration
```json
{
"version": "1.0.0",
"initialized": true,
"created": "2025-11-02T12:00:00Z",
"excluded_patterns": [
"node_modules/**",
".git/**",
"dist/**",
"build/**",
".next/**",
"__pycache__/**",
"*.pyc",
"*.log"
],
"focus_areas": [
"architecture",
"security",
"performance",
"testing"
],
"output_format": "markdown",
"parallel_execution": true,
"incremental_updates": true
}
```
## Configuration Options
### Excluded Patterns
**Files/directories to skip**:
```json
"excluded_patterns": [
"node_modules/**", // Dependencies
".git/**", // Version control
"dist/**", "build/**", // Build outputs
"coverage/**", // Test coverage
"*.min.js", // Minified files
"vendor/**" // Third-party code
]
```
### Focus Areas
**Analysis priorities**:
```json
"focus_areas": [
"architecture", // System design
"security", // Vulnerabilities
"performance", // Bottlenecks
"testing", // Test coverage
"documentation" // Code docs
]
```
### Execution Options
```json
"parallel_execution": true, // Run agents in parallel (55% faster)
"incremental_updates": true // Enable delta updates
```
## Common Customizations
### For Large Monorepos
```json
{
"excluded_patterns": [
"packages/*/node_modules/**",
"apps/*/dist/**",
"*.lock"
],
"parallel_execution": true
}
```
### For Security-Focused Analysis
```json
{
"focus_areas": ["security", "quality"],
"deep_scan_enabled": true
}
```
### For Fast Iterations
```json
{
"excluded_patterns": [
"**/*.test.ts",
"**/*.spec.ts",
"**/__tests__/**"
],
"parallel_execution": true
}
```
## Validation
After editing, validate config:
```bash
jq empty .claude/steering/config.json && echo "✓ Valid JSON" || echo "✗ Invalid JSON"
```
---
**Customize your analysis:** Edit `.claude/steering/config.json`

137
commands/steering-export.md Normal file
View File

@@ -0,0 +1,137 @@
---
description: Export steering context to different formats (JSON, YAML, HTML, PDF - Coming Soon)
---
# Steering Context Generator - Export
Export generated documentation to different formats.
## Quick Start
```bash
/steering-export --format json
```
## Supported Formats (v1.0)
### Markdown (Default)
Already generated in `.claude/steering/*.md`
### JSON
Export as structured JSON:
```bash
# Export all documents
cat > .claude/steering/export/context.json << 'EOF'
{
"version": "1.0.0",
"generated": "$(date -Iseconds)",
"project": {
"name": "$(basename $(pwd))",
"tech_stack": "$(jq -r '.tech_stack' .claude/memory/orchestration/detection.json 2>/dev/null || echo 'Unknown')"
},
"documents": {
"architecture": "$(cat .claude/steering/ARCHITECTURE.md | base64)",
"ai_context": "$(cat .claude/steering/AI_CONTEXT.md | base64)",
"codebase_guide": "$(cat .claude/steering/CODEBASE_GUIDE.md | base64)"
}
}
EOF
echo "Exported to: .claude/steering/export/context.json"
```
### Plain Text
Strip markdown formatting:
```bash
# Convert markdown to plain text
for file in .claude/steering/*.md; do
BASENAME=$(basename "$file" .md)
pandoc "$file" -t plain -o ".claude/steering/export/${BASENAME}.txt" 2>/dev/null || \
sed 's/[#*`]//g' "$file" > ".claude/steering/export/${BASENAME}.txt"
done
echo "Exported to: .claude/steering/export/*.txt"
```
## Coming Soon (v1.1+)
### HTML
```bash
# Will generate interactive HTML documentation
/steering-export --format html
```
### PDF
```bash
# Will generate PDF documentation
/steering-export --format pdf
```
### YAML
```bash
# Will generate YAML configuration
/steering-export --format yaml
```
## Export Directory
Exports are saved to:
```
.claude/steering/export/
├── context.json
├── ARCHITECTURE.txt
├── AI_CONTEXT.txt
└── ...
```
## Use Cases
### Share with Team
```bash
# Export and compress
/steering-export --format json
tar -czf context-export.tar.gz .claude/steering/export/
```
### CI/CD Integration
```bash
# Export as JSON for automated processing
/steering-export --format json
cat .claude/steering/export/context.json | jq '.documents.architecture'
```
### Documentation Website
```bash
# Convert to HTML (v1.1+)
/steering-export --format html
# Publish to docs site
```
## Troubleshooting
### Export fails
Ensure documents are generated:
```bash
/steering-status
```
If missing, generate first:
```bash
/steering-generate
```
---
**Share your context:** Export with `/steering-export`

View File

@@ -0,0 +1,432 @@
---
description: Generate comprehensive steering context documentation by analyzing your codebase with specialized AI agents
---
# Steering Context Generator - Full Generation
Analyze your entire codebase and generate comprehensive AI-readable documentation.
## Quick Start
```bash
/steering-generate
```
That's it! The system will:
1. 🔍 Detect your project type and tech stack
2. 📊 Assess complexity and select workflow
3. 🤖 Execute specialized agents in parallel
4. 📝 Generate comprehensive documentation
5. ✅ Validate and save outputs
## What Gets Generated
The following documents are created in `.claude/steering/`:
### Core Documents (Always Generated)
| Document | Purpose | Typical Size |
|----------|---------|--------------|
| `ARCHITECTURE.md` | System architecture, components, data flow | 200-400 KB |
| `AI_CONTEXT.md` | Bootstrap context for AI agents | 100-200 KB |
| `CODEBASE_GUIDE.md` | Developer onboarding guide | 150-300 KB |
### Extended Documents (Based on Project Type)
| Document | Generated When | Purpose |
|----------|----------------|---------|
| `DOMAIN_CONTEXT.md` | Complex projects | Business logic and rules |
| `QUALITY_REPORT.md` | Always | Security, performance analysis |
| `UI_DESIGN_SYSTEM.md` | Frontend detected | Component catalog, design tokens |
| `TESTING_GUIDE.md` | Tests found | Testing patterns, coverage |
| `DATABASE_CONTEXT.md` | Database detected | Schema, DAL patterns |
| `MESSAGING_GUIDE.md` | Queues/events found | Event catalog, pub/sub |
| `API_DESIGN_GUIDE.md` | API endpoints found | REST standards, error handling |
| `STRIPE_PAYMENT_CONTEXT.md` | Stripe integration found | Payment flows, webhook handlers, PCI compliance |
| `AUTH0_OAUTH_CONTEXT.md` | Auth0 integration found | OAuth flows, configuration, security assessment |
| `PAYLOAD_CMS_CONTEXT.md` | Payload CMS detected | CMS architecture, content models, API configuration |
| `PAYLOAD_CMS_CONFIG.md` | Payload CMS detected | Configuration analysis, security audit, compliance |
| `DESIGN_SYSTEM_ARCHITECTURE.md` | Design tokens/components detected | Design token analysis, component library structure, maturity |
| `UI_FRAMEWORK_GUIDE.md` | UI framework detected (React, Vue, etc.) | Framework configuration, component patterns, best practices |
| `WEB_UI_DESIGN_CONTEXT.md` | Frontend pages/components detected | Web UI design analysis, accessibility, UX flows, responsive design |
## How It Works
### Phase 1: Project Detection
The system automatically detects:
**Tech Stack**:
- Package managers (npm, pnpm, pip, cargo, go, maven, gradle)
- Frameworks (Next.js, React, Django, FastAPI, etc.)
- Databases (Prisma, Drizzle, TypeORM, MongoDB, etc.)
- Testing frameworks (Jest, Vitest, pytest, etc.)
- **UI frameworks** (React, Vue, Angular, Svelte - if detected)
- **Design system tools** (Tailwind, Shadcn, Storybook, design tokens)
- **Auth0 OAuth integration** (if @auth0 SDK detected)
- **Payload CMS integration** (if @payloadcms packages detected)
**Project Structure**:
- Monorepo vs single-package
- Microservices vs monolith
- Frontend, backend, or full-stack
- File count and directory depth
**Complexity Assessment**:
```
Simple: < 50 files, < 3 levels deep → 20 min
Moderate: 50-200 files, 3-6 levels deep → 45 min
Complex: 200+ files, 6+ levels deep → 85 min
```
### Phase 2: Agent Selection
Based on detection, the system selects appropriate agents:
**Foundation Agents** (Always Run):
- `structure-analyst`: Map file system and dependencies
- `pattern-detective`: Identify architectural patterns
- `quality-auditor`: Security and quality analysis
**Domain-Specific Agents** (Conditional):
- `ui-specialist`: If frontend components found
- **`design-system-architect`**: If design tokens/component library detected
- **`ui-framework-analyzer`**: If UI framework detected (React, Vue, Angular, Svelte)
- **`web-ui-design-analyzer`**: If frontend pages/components found
- `test-strategist`: If test files found
- `database-analyst`: If database schemas found
- `messaging-architect`: If queues/events found
- `api-design-analyst`: If API routes found
- **`auth0-detector`**: If Auth0 SDK imports or configuration found
- **`oauth-security-auditor`**: If Auth0 integration found (runs after auth0-detector)
- **`payload-cms-detector`**: If Payload CMS packages detected
- **`payload-cms-config-analyzer`**: If Payload CMS detected (runs after payload-cms-detector)
**Synthesis Agent** (Always Final):
- `context-synthesizer`: Generate final documentation
### Phase 3: Parallel Execution
Agents execute in intelligent parallel groups:
```mermaid
Group 1 (Foundation):
structure-analyst ───────┐
integration-mapper ──────┤ Run in parallel
ui-specialist ───────────┘
Group 2 (Analysis) - Depends on Group 1:
domain-expert ──────────┐
pattern-detective ──────┤
test-strategist ────────┤ Run in parallel
database-analyst ───────┤
design-system-architect ┘
Group 3 (UI/Framework/Design) - Depends on Groups 1 & 2:
ui-framework-analyzer ──┐
web-ui-design-analyzer ─┤ Run in parallel
messaging-architect ────┤
api-design-analyst ─────┤
stripe-payment-expert ──├ Run in parallel
auth0-detector ─────────┤
payload-cms-detector ───┤
quality-auditor ────────┘
Group 3B (Security & Config Audits) - Depends on Group 3:
oauth-security-auditor (sequential, after auth0-detector, if Auth0 detected)
payload-cms-config-analyzer (sequential, after payload-cms-detector, if Payload CMS detected)
Group 4 (Synthesis) - Depends on all:
context-synthesizer (sequential)
```
**Time Savings**: Parallel execution is 55% faster than sequential!
**Note**: UI/Framework/Design analysis adds ~30-40 minutes for comprehensive analysis if UI detected.
**Note**: Auth0 security audit runs automatically after Auth0 detection, adding ~10 minutes if Auth0 is present.
**Note**: Payload CMS config analysis runs automatically after CMS detection, adding ~10 minutes if Payload CMS is present.
### Phase 4: Output Generation
Each agent contributes to final documents:
```
structure-analyst → ARCHITECTURE.md (structure section)
domain-expert → DOMAIN_CONTEXT.md (complete)
pattern-detective → ARCHITECTURE.md (patterns section)
ui-specialist → UI_DESIGN_SYSTEM.md (complete)
design-system-architect → DESIGN_SYSTEM_ARCHITECTURE.md (complete, if design system found)
ui-framework-analyzer → UI_FRAMEWORK_GUIDE.md (complete, if UI framework found)
web-ui-design-analyzer → WEB_UI_DESIGN_CONTEXT.md (complete, if frontend pages found)
test-strategist → TESTING_GUIDE.md (complete)
database-analyst → DATABASE_CONTEXT.md (complete)
messaging-architect → MESSAGING_GUIDE.md (complete)
api-design-analyst → API_DESIGN_GUIDE.md (complete)
stripe-payment-expert → STRIPE_PAYMENT_CONTEXT.md (complete, if Stripe found)
auth0-detector → AUTH0_OAUTH_CONTEXT.md (complete, if Auth0 found)
oauth-security-auditor → AUTH0_SECURITY_AUDIT.md (complete, if Auth0 found)
payload-cms-detector → PAYLOAD_CMS_CONTEXT.md (complete, if Payload CMS found)
payload-cms-config-analyzer → PAYLOAD_CMS_CONFIG.md (complete, if Payload CMS found)
quality-auditor → QUALITY_REPORT.md (complete)
context-synthesizer → AI_CONTEXT.md, CODEBASE_GUIDE.md
```
## Execution Workflow
The system uses the Task tool to invoke agents in parallel:
### Step 1: Initialize Session
```bash
# Create session ID for tracking
SESSION_ID="gen_$(date +%Y%m%d_%H%M%S)"
# Initialize execution state
cat > .claude/memory/orchestration/current_session.json << EOF
{
"session_id": "$SESSION_ID",
"started": "$(date -Iseconds)",
"status": "running",
"phase": "detection"
}
EOF
```
### Step 2: Detect and Assess
Analyze project characteristics:
```markdown
Detecting project type...
✓ Tech Stack: Node.js/TypeScript
✓ Framework: Next.js 14 (App Router)
✓ Package Manager: pnpm (monorepo detected)
✓ Database: Prisma + PostgreSQL
✓ Testing: Vitest + Playwright
✓ CMS: Payload CMS v2.x detected
Assessing complexity...
Files: 387
Directories: 45 (max depth: 6)
Dependencies: 73
Estimated LOC: ~25,000
Complexity: Moderate
Estimated Time: 55 minutes (includes Payload CMS analysis)
Workflow: Standard (3 parallel phases + security audits)
```
### Step 3: Execute Foundation Agents (Group 1)
Use the Task tool to run agents in parallel:
**Critical**: Execute ALL agents in Group 1 in a SINGLE message with multiple Task tool calls.
### Step 4: Execute Analysis Agents (Group 2)
**Depends on**: Group 1 outputs
### Step 5: Execute Architecture Agents (Group 3)
**Depends on**: Groups 1 & 2 outputs
### Step 5B: Execute Security Audits (Group 3B, Sequential)
**Depends on**: Group 3 outputs
Automatically executes after:
- Auth0 detection (if Auth0 found)
- Payload CMS detection (if Payload CMS found)
### Step 6: Final Synthesis (Group 4)
**Depends on**: All previous outputs
### Step 7: Validation and Completion
```bash
# Validate generated files
bash scripts/validate.sh
# Display summary
echo "✅ Steering context generation complete!"
echo ""
echo "Generated Files (.claude/steering/):"
ls -lh .claude/steering/*.md | awk '{print " ✓", $9, "(" $5 ")"}'
echo ""
echo "Next Steps:"
echo " 1. Review: /steering-status"
echo " 2. Load context: Reference .claude/steering/*.md in prompts"
echo " 3. Update later: /steering-update (incremental)"
```
## Configuration Options
### Focus Areas
Prioritize specific analysis areas in `.claude/steering/config.json`:
```json
{
"focus_areas": ["architecture", "security", "performance"]
}
```
### Excluded Patterns
Skip certain files/directories:
```json
{
"excluded_patterns": [
"node_modules/**",
".git/**",
"dist/**",
"*.test.ts"
]
}
```
### Parallel Execution
Disable parallel execution if needed:
```json
{
"parallel_execution": false
}
```
## Expected Output
After successful completion:
```
✅ Steering context generation complete! (55 minutes)
Generated Files (.claude/steering/):
✓ ARCHITECTURE.md (342 KB) - System architecture
✓ AI_CONTEXT.md (156 KB) - AI agent bootstrap
✓ CODEBASE_GUIDE.md (278 KB) - Developer guide
✓ DOMAIN_CONTEXT.md (189 KB) - Business logic
✓ QUALITY_REPORT.md (134 KB) - Quality analysis
✓ PAYLOAD_CMS_CONTEXT.md (203 KB) - CMS architecture
✓ PAYLOAD_CMS_CONFIG.md (187 KB) - CMS configuration
Total: 2.1 MB of context documentation
Performance:
Total tokens: ~165,000
Agents executed: 16
Parallel efficiency: 52% time saved
Next Steps:
1. Review: /steering-status
2. Load context: Reference .claude/steering/*.md in prompts
3. Update later: /steering-update (incremental)
```
## Troubleshooting
### Generation Interrupted
If generation is interrupted:
```bash
# Check for checkpoint
ls .claude/memory/orchestration/current_session.json
# Resume from checkpoint
/steering-resume
```
### Agents Not Found
Ensure plugin is properly installed:
```bash
/plugin list
# Should show "steering-context-generator"
# Reinstall if needed
/plugin uninstall steering-context-generator@aditi.code
/plugin install steering-context-generator@aditi.code
```
### Out of Memory
For very large codebases:
1. Disable parallel execution
2. Run selective analysis
3. Increase excluded patterns
4. Use incremental approach
## Advanced Usage
### Selective Analysis
Analyze only specific areas:
```json
{
"focus_areas": ["cms", "api"],
"agents": ["payload-cms-detector", "api-design-analyst"]
}
```
### Multi-Project Analysis
Analyze multiple projects:
```bash
# Project A
cd /path/to/project-a
/steering-generate
# Project B
cd /path/to/project-b
/steering-generate
# Compare
diff .claude/steering/ARCHITECTURE.md ../project-a/.claude/steering/ARCHITECTURE.md
```
## Performance Tips
**Faster Generation**:
- ✅ Use parallel execution (enabled by default)
- ✅ Exclude large generated directories
- ✅ Use Haiku model for simple projects
- ✅ Enable incremental updates
**Better Quality**:
- ✅ Use Sonnet model (default)
- ✅ Use Opus model for complex analysis
- ✅ Provide clear focus areas
- ✅ Review and refine outputs
## Success Metrics
Your generation is successful when:
- ✅ All expected files generated
- ✅ Files are valid markdown
- ✅ File sizes are reasonable (not empty, not too large)
- ✅ Validation passes
- ✅ Content is relevant and accurate
## Next Steps
After generation:
1. **Review Output**: `/steering-status`
2. **Load Context**: Reference `.claude/steering/*.md` in AI prompts
3. **Incremental Updates**: `/steering-update` when code changes
4. **Export**: `/steering-export` for different formats
5. **Clean Up**: `/steering-clean` to remove old archives
---
**Ready to generate?** Just run: `/steering-generate`
The system handles everything automatically!

136
commands/steering-resume.md Normal file
View File

@@ -0,0 +1,136 @@
---
description: Resume an interrupted generation from the last checkpoint
---
# Steering Context Generator - Resume
Continue a generation that was interrupted or failed.
## Quick Start
```bash
/steering-resume
```
## When to Use
- ⚠ Generation was interrupted (Ctrl+C, timeout, crash)
- ⚠ Agent execution failed mid-way
- ⚠ System resources exhausted
- ⚠ Network connectivity issues
## How It Works
### Checkpoint System
The system automatically saves checkpoints:
```
.claude/memory/orchestration/
├── current_session.json # Current execution state
├── checkpoint_group1.json # After Group 1 completes
├── checkpoint_group2.json # After Group 2 completes
└── checkpoint_group3.json # After Group 3 completes
```
### Resume Logic
```bash
# Check for checkpoint
if [ -f ".claude/memory/orchestration/current_session.json" ]; then
PHASE=$(jq -r '.phase' .claude/memory/orchestration/current_session.json)
STATUS=$(jq -r '.status' .claude/memory/orchestration/current_session.json)
if [ "$STATUS" == "running" ] || [ "$STATUS" == "failed" ]; then
echo "Found interrupted session at phase: $PHASE"
echo "Resuming from checkpoint..."
# Resume from last completed phase
case $PHASE in
"group1")
# Restart Group 1 or skip if completed
;;
"group2")
# Skip Group 1, resume Group 2
;;
"group3")
# Skip Groups 1 & 2, resume Group 3
;;
"synthesis")
# Run final synthesis only
;;
esac
fi
fi
```
## Expected Output
```
🔄 Checking for interrupted session...
Found Session: gen_20251102_120000
Status: interrupted
Phase: group2 (60% complete)
Last Activity: 2025-11-02 12:45:00
Agents Completed:
✓ structure-analyst (Group 1)
✓ integration-mapper (Group 1)
✓ ui-specialist (Group 1)
✓ domain-expert (Group 2)
⏳ pattern-detective (Group 2) - INTERRUPTED
Resuming from Group 2...
────────────────────────────────────────
Phase 2/3: Deep Analysis (Resumed)
✓ domain-expert: Already complete
⏳ pattern-detective: Resuming analysis...
⏳ test-strategist: Starting fresh...
⏳ database-analyst: Starting fresh...
[Continue with normal execution...]
```
## Manual Resume
If automatic resume fails:
```bash
# Check what's completed
ls .claude/memory/*/
# Manually continue with remaining agents
# Use Task tool to invoke specific agents that didn't complete
```
## Clearing Failed State
To start fresh (discard interrupted session):
```bash
# Remove checkpoint
rm .claude/memory/orchestration/current_session.json
# Run full generation
/steering-generate
```
## Troubleshooting
### "No checkpoint found"
The session completed or never started. Run `/steering-generate`.
### Resume keeps failing
Clear state and start fresh:
```bash
rm .claude/memory/orchestration/*.json
/steering-setup
/steering-generate
```
---
**Recover from interruptions:** Run `/steering-resume`

289
commands/steering-status.md Normal file
View File

@@ -0,0 +1,289 @@
---
description: Check the status of Steering Context Generator installation, last generation, and generated files
---
# Steering Context Generator - Status
View system status, generated files, and recommendations.
## Quick Start
```bash
/steering-status
```
## Expected Output
```
📊 Steering Context Generator - Status
═══════════════════════════════════════════════════════════
INSTALLATION
Plugin Version: 1.0.0
Installed: 2025-11-01 14:00:00
Status: ✓ Ready
CONFIGURATION
Config File: .claude/steering/config.json
Parallel Execution: ✓ Enabled
Incremental Updates: ✓ Enabled
Output Format: markdown
LAST GENERATION
Date: 2025-11-01 15:30:45 (1 day ago)
Duration: 44 minutes
Workflow: Standard (Moderate complexity)
Agents: 7 executed
Status: ✓ Complete
GENERATED FILES (.claude/steering/)
✓ ARCHITECTURE.md 342 KB Fresh
✓ AI_CONTEXT.md 156 KB Fresh
✓ CODEBASE_GUIDE.md 278 KB Fresh
✓ DOMAIN_CONTEXT.md 189 KB Fresh
✓ QUALITY_REPORT.md 134 KB Fresh
✓ TESTING_GUIDE.md 167 KB Fresh
✓ UI_DESIGN_SYSTEM.md 203 KB Fresh
MEMORY USAGE
Total: 1.2 MB
Structure: 127 KB
Domain: 189 KB
Patterns: 98 KB
Quality: 134 KB
UI: 203 KB
Testing: 167 KB
Archives: 450 KB (2 previous runs)
CODEBASE STATUS
Files Tracked: 387
Last Change: 1 day ago (12 files modified)
Tech Stack: Node.js/TypeScript, Next.js 14
Complexity: Moderate
═══════════════════════════════════════════════════════════
RECOMMENDATIONS
⚠ Context may be outdated (12 files changed)
→ Run: /steering-update
💡 Archive size growing (450 KB)
→ Run: /steering-clean
═══════════════════════════════════════════════════════════
QUICK ACTIONS
/steering-update Update with latest changes
/steering-generate Full regeneration
/steering-clean Clean up archives
/steering-export Export to other formats
```
## Implementation
```bash
#!/bin/bash
echo "📊 Steering Context Generator - Status"
echo ""
echo "═══════════════════════════════════════════════════════════"
echo ""
# Check installation
if [ ! -f ".claude/steering/config.json" ]; then
echo "❌ NOT INSTALLED"
echo ""
echo "Run: /steering-setup"
exit 1
fi
# Plugin version
echo "INSTALLATION"
echo " Plugin Version: 1.0.0"
if [ -f ".claude/memory/orchestration/state.json" ]; then
INSTALLED=$(jq -r '.timestamp' .claude/memory/orchestration/state.json)
echo " Installed: $INSTALLED"
echo " Status: ✓ Ready"
else
echo " Status: ⚠ Incomplete setup"
fi
echo ""
# Configuration
echo "CONFIGURATION"
echo " Config File: .claude/steering/config.json"
PARALLEL=$(jq -r '.parallel_execution' .claude/steering/config.json)
INCREMENTAL=$(jq -r '.incremental_updates' .claude/steering/config.json)
FORMAT=$(jq -r '.output_format' .claude/steering/config.json)
echo " Parallel Execution: $([ "$PARALLEL" == "true" ] && echo "✓ Enabled" || echo "✗ Disabled")"
echo " Incremental Updates: $([ "$INCREMENTAL" == "true" ] && echo "✓ Enabled" || echo "✗ Disabled")"
echo " Output Format: $FORMAT"
echo ""
# Last generation
echo "LAST GENERATION"
if [ -f ".claude/memory/orchestration/state.json" ]; then
LAST_RUN=$(jq -r '.last_run' .claude/memory/orchestration/state.json)
if [ "$LAST_RUN" != "null" ]; then
echo " Date: $LAST_RUN"
echo " Status: ✓ Complete"
else
echo " Status: ⚠ Never run"
echo " → Run: /steering-generate"
fi
else
echo " Status: ⚠ No generation data"
fi
echo ""
# Generated files
echo "GENERATED FILES (.claude/steering/)"
if [ -d ".claude/steering" ]; then
for file in .claude/steering/*.md; do
if [ -f "$file" ]; then
BASENAME=$(basename "$file")
SIZE=$(du -h "$file" | cut -f1)
AGE=$(find "$file" -mtime -1 2>/dev/null && echo "Fresh" || echo "Stale")
echo "$BASENAME$(printf '%*s' $((30-${#BASENAME})) '')$SIZE $AGE"
fi
done
else
echo " ⚠ No files generated"
fi
echo ""
# Memory usage
echo "MEMORY USAGE"
if [ -d ".claude/memory" ]; then
TOTAL=$(du -sh .claude 2>/dev/null | cut -f1)
echo " Total: $TOTAL"
else
echo " No memory data"
fi
echo ""
# Codebase status
echo "CODEBASE STATUS"
FILE_COUNT=$(find . -type f \
-not -path "*/node_modules/*" \
-not -path "*/.git/*" \
-not -path "*/dist/*" \
2>/dev/null | wc -l | tr -d ' ')
echo " Files Tracked: $FILE_COUNT"
if git rev-parse --git-dir > /dev/null 2>&1; then
LAST_COMMIT=$(git log -1 --format=%cd --date=relative)
echo " Last Change: $LAST_COMMIT"
fi
echo ""
echo "═══════════════════════════════════════════════════════════"
echo ""
# Recommendations
echo "RECOMMENDATIONS"
# Check if update needed
if [ -f ".claude/steering/ARCHITECTURE.md" ]; then
CHANGED=$(find . -newer .claude/steering/ARCHITECTURE.md -type f \
-not -path "*/node_modules/*" \
-not -path "*/.git/*" \
2>/dev/null | wc -l | tr -d ' ')
if [ "$CHANGED" -gt 10 ]; then
echo " ⚠ Context may be outdated ($CHANGED files changed)"
echo " → Run: /steering-update"
echo ""
fi
fi
# Check archive size
if [ -d ".claude/memory/archives" ]; then
ARCHIVE_SIZE=$(du -sm .claude/memory/archives 2>/dev/null | cut -f1)
if [ "$ARCHIVE_SIZE" -gt 100 ]; then
echo " 💡 Archive size growing (${ARCHIVE_SIZE}MB)"
echo " → Run: /steering-clean"
echo ""
fi
fi
echo "═══════════════════════════════════════════════════════════"
echo ""
echo "QUICK ACTIONS"
echo " /steering-update Update with latest changes"
echo " /steering-generate Full regeneration"
echo " /steering-clean Clean up archives"
echo " /steering-export Export to other formats"
```
## Status Indicators
**Installation Status**:
- ✓ Ready: Fully installed and configured
- ⚠ Incomplete: Missing files or configuration
- ❌ Not Installed: Run `/steering-setup`
**Generation Status**:
- ✓ Complete: Successfully generated
- ⏳ Running: Generation in progress
- ⚠ Never run: No generation yet
- ❌ Failed: Last generation failed
**File Freshness**:
- Fresh: Modified within 24 hours
- Stale: Modified >24 hours ago
- Missing: Expected file not found
## Checking Specific Information
### Installation Details
```bash
cat .claude/steering/config.json | jq '.'
```
### Generation History
```bash
cat .claude/memory/orchestration/state.json | jq '.'
```
### Memory Breakdown
```bash
du -sh .claude/memory/*/ | sort -h
```
### File Details
```bash
ls -lh .claude/steering/*.md
```
### Agent Status
```bash
cat .claude/memory/orchestration/agents.json | jq '.agents'
```
## Troubleshooting
### "Not Installed" Message
```bash
/steering-setup
```
### Missing Files
```bash
# Regenerate
/steering-generate
# Or validate
bash scripts/validate.sh
```
### Incorrect Status
```bash
# Reset state
rm .claude/memory/orchestration/state.json
bash scripts/init.sh
```
---
**Quick check:** Run `/steering-status` anytime!

457
commands/steering-update.md Normal file
View File

@@ -0,0 +1,457 @@
---
description: Incrementally update steering context based on code changes since last generation (80% faster than full regeneration)
---
# Steering Context Generator - Incremental Update
Efficiently update your steering context documentation when code changes.
## Quick Start
```bash
/steering-update
```
The system will:
1. 🔍 Detect changed files since last generation
2. 📊 Identify affected domains
3. 🤖 Run only relevant agents
4. 📝 Update specific documents
5. ✅ Merge with existing context
## How It Works
### Change Detection
The system uses Git (if available) or file timestamps to detect changes:
```bash
# Check last generation time
LAST_GEN=$(jq -r '.last_run' .claude/memory/orchestration/state.json)
# Detect changes via Git
git diff --name-only $LAST_GEN..HEAD
# Or via file timestamps
find . -newer .claude/steering/ARCHITECTURE.md -type f
```
### Domain Mapping
Changed files are mapped to affected domains:
| Files Changed | Affected Domains | Agents to Run |
|---------------|------------------|---------------|
| `*.tsx, *.jsx` | UI Components | `ui-specialist` |
| `app/api/*.ts` | API Routes | `api-design-analyst` |
| `prisma/schema.prisma` | Database | `database-analyst` |
| `*.test.ts` | Testing | `test-strategist` |
| `lib/events/*.ts` | Messaging | `messaging-architect` |
### Selective Agent Execution
Only affected agents run:
```
Changes:
✓ UI: 4 files modified
✓ API: 6 files changed
✓ Database: 2 schema updates
Running agents:
⏳ ui-specialist (updating UI_DESIGN_SYSTEM.md)
⏳ api-design-analyst (updating API_DESIGN_GUIDE.md)
⏳ database-analyst (updating DATABASE_CONTEXT.md)
```
### Document Merging
Updated sections are merged with existing documents:
```markdown
Before:
UI_DESIGN_SYSTEM.md (203 KB, 45 components)
Changes:
+ 2 new components
~ 3 modified components
After:
UI_DESIGN_SYSTEM.md (237 KB, 47 components)
```
## Execution Workflow
### Step 1: Detect Changes
```bash
# Load last generation timestamp
LAST_RUN=$(jq -r '.last_run' .claude/memory/orchestration/state.json)
if [ "$LAST_RUN" == "null" ]; then
echo "❌ No previous generation found. Run /steering-generate first."
exit 1
fi
# Detect changed files
echo "🔍 Detecting changes since $LAST_RUN..."
if git rev-parse --git-dir > /dev/null 2>&1; then
# Use Git if available
CHANGED_FILES=$(git diff --name-only $LAST_RUN..HEAD)
else
# Use file timestamps
CHANGED_FILES=$(find . -newer .claude/steering/ARCHITECTURE.md -type f \
-not -path "*/node_modules/*" \
-not -path "*/.git/*")
fi
if [ -z "$CHANGED_FILES" ]; then
echo "✓ No changes detected. Context is up-to-date."
exit 0
fi
echo "Found $(echo "$CHANGED_FILES" | wc -l) changed files"
```
### Step 2: Analyze Change Scope
```bash
# Categorize changes
UI_CHANGES=$(echo "$CHANGED_FILES" | grep -E '\.(tsx|jsx|css|scss)$' | wc -l)
API_CHANGES=$(echo "$CHANGED_FILES" | grep -E 'api/.*\.(ts|js)$' | wc -l)
DB_CHANGES=$(echo "$CHANGED_FILES" | grep -E '(schema|migration)' | wc -l)
TEST_CHANGES=$(echo "$CHANGED_FILES" | grep -E '\.(test|spec)\.' | wc -l)
echo "Change analysis:"
echo " UI components: $UI_CHANGES files"
echo " API routes: $API_CHANGES files"
echo " Database: $DB_CHANGES files"
echo " Tests: $TEST_CHANGES files"
```
### Step 3: Select Agents
```bash
AGENTS_TO_RUN=()
if [ $UI_CHANGES -gt 0 ]; then
AGENTS_TO_RUN+=("ui-specialist")
fi
if [ $API_CHANGES -gt 0 ]; then
AGENTS_TO_RUN+=("api-design-analyst")
fi
if [ $DB_CHANGES -gt 0 ]; then
AGENTS_TO_RUN+=("database-analyst")
fi
if [ $TEST_CHANGES -gt 0 ]; then
AGENTS_TO_RUN+=("test-strategist")
fi
# Always run quality auditor for affected areas
AGENTS_TO_RUN+=("quality-auditor")
echo "Selected agents: ${AGENTS_TO_RUN[@]}"
```
### Step 4: Execute Agents in Parallel
Use the Task tool to run selected agents:
```markdown
For UI changes - Invoke ui-specialist:
Update UI_DESIGN_SYSTEM.md with new/modified components
Focus only on changed files: $UI_CHANGED_FILES
Merge with existing: .claude/memory/ui/
For API changes - Invoke api-design-analyst:
Update API_DESIGN_GUIDE.md with new/modified endpoints
Focus only on changed files: $API_CHANGED_FILES
Merge with existing: .claude/memory/api-design/
For Database changes - Invoke database-analyst:
Update DATABASE_CONTEXT.md with schema changes
Focus only on migrations/schema: $DB_CHANGED_FILES
Merge with existing: .claude/memory/database/
For Test changes - Invoke test-strategist:
Update TESTING_GUIDE.md with new test patterns
Focus only on changed tests: $TEST_CHANGED_FILES
Merge with existing: .claude/memory/testing/
```
### Step 5: Validate and Merge
```bash
# Validate updated documents
bash scripts/validate.sh
# Update architecture document with changes
# (context-synthesizer runs lightweight merge)
# Update state
cat > .claude/memory/orchestration/state.json << EOF
{
"phase": "ready",
"timestamp": "$(date -Iseconds)",
"initialized": true,
"last_run": "$(date -Iseconds)",
"last_update": "$(date -Iseconds)",
"agents_status": {
$(for agent in "${AGENTS_TO_RUN[@]}"; do
echo "\"$agent\": \"updated\","
done | sed '$ s/,$//')
}
}
EOF
```
### Step 6: Display Summary
```bash
echo "✅ Update complete!"
echo ""
echo "Updated Files:"
for agent in "${AGENTS_TO_RUN[@]}"; do
case $agent in
ui-specialist)
echo " ↻ UI_DESIGN_SYSTEM.md (+${UI_CHANGES} changes)"
;;
api-design-analyst)
echo " ↻ API_DESIGN_GUIDE.md (+${API_CHANGES} changes)"
;;
database-analyst)
echo " ↻ DATABASE_CONTEXT.md (+${DB_CHANGES} changes)"
;;
test-strategist)
echo " ↻ TESTING_GUIDE.md (+${TEST_CHANGES} changes)"
;;
quality-auditor)
echo " ↻ QUALITY_REPORT.md (revalidated)"
;;
esac
done
echo ""
echo "Time saved vs full regeneration: ~$(echo "$((45 - 8))" minutes)"
echo "Tokens used: ~18,000 (vs ~145,000 full)"
```
## Expected Output
```
🔄 Checking for changes since last generation...
Last Generated: 2025-11-01 15:30:45 (1 day ago)
Changes Detected (git diff):
Modified: 12 files
Added: 3 files
Deleted: 1 file
Change Analysis:
UI components: 4 files
API routes: 6 files
Database schema: 2 files
Selected Agents:
✓ ui-specialist
✓ api-design-analyst
✓ database-analyst
✓ quality-auditor
────────────────────────────────────────
Running Incremental Analysis:
⏳ ui-specialist: Updating UI_DESIGN_SYSTEM.md...
⏳ api-design-analyst: Updating API_DESIGN_GUIDE.md...
⏳ database-analyst: Updating DATABASE_CONTEXT.md...
⏳ quality-auditor: Revalidating affected areas...
────────────────────────────────────────
✅ Update complete! (8 minutes)
Updated Files:
↻ UI_DESIGN_SYSTEM.md (+34 KB, 2 new components)
↻ API_DESIGN_GUIDE.md (+12 KB, 3 endpoints modified)
↻ DATABASE_CONTEXT.md (+8 KB, 1 table added)
✓ ARCHITECTURE.md (revalidated)
✓ QUALITY_REPORT.md (revalidated)
Performance:
Time: 8 minutes (vs 45 full)
Time saved: 37 minutes (82% faster)
Tokens: ~18,000 (vs ~145,000 full)
Efficiency: 87% token savings
Next Steps:
/steering-status - View updated context
```
## Configuration
### Update Threshold
Set minimum changes to trigger update:
```json
// .claude/steering/config.json
{
"update_threshold": 5 // Minimum files changed
}
```
### Force Full Regeneration
Skip incremental and force full regeneration:
```bash
/steering-generate --force
```
### Selective Update
Update only specific domains:
```bash
# Update only UI (not yet implemented - use config)
{
"update_domains": ["ui"]
}
```
## Troubleshooting
### "No changes detected" but files changed
**Causes**:
1. Files in excluded patterns
2. Git not tracking changes
3. Files outside analysis scope
**Solutions**:
```bash
# Check excluded patterns
cat .claude/steering/config.json | jq '.excluded_patterns'
# Force full regeneration
/steering-generate --force
# Update excluded patterns if needed
```
### Merge conflicts
If updates conflict with manual edits:
**Option 1**: Backup and regenerate
```bash
cp .claude/steering/ARCHITECTURE.md .claude/steering/ARCHITECTURE.md.backup
/steering-update
# Manually merge if needed
```
**Option 2**: Force full regeneration
```bash
/steering-generate --force
```
### Agent execution fails
Check agent status:
```bash
cat .claude/memory/orchestration/state.json | jq '.agents_status'
```
Retry specific agent:
```bash
# Run agent manually with Task tool
```
## Best Practices
**When to Use Update**:
- ✅ Small to medium changes (<20% of files)
- ✅ Focused changes in specific domains
- ✅ Regular maintenance updates
- ✅ After feature additions
**When to Use Full Regeneration**:
- ✅ Major refactoring (>20% of files)
- ✅ Architecture changes
- ✅ First-time generation
- ✅ After long periods without updates
**Update Frequency**:
- Daily: For active development
- Weekly: For stable projects
- After features: After completing features
- Before releases: Ensure docs current
## Advanced Usage
### Automated Updates
Set up Git hooks for automatic updates:
```bash
# .git/hooks/post-commit
#!/bin/bash
if [ -f ".claude/steering/config.json" ]; then
/steering-update
fi
```
### CI/CD Integration
Add to CI pipeline:
```yaml
# .github/workflows/update-context.yml
name: Update Steering Context
on:
push:
branches: [main]
jobs:
update:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Update Context
run: claude /steering-update
```
### Diff Analysis
Compare before/after:
```bash
# Before update
cp .claude/steering/ARCHITECTURE.md /tmp/arch-before.md
# Run update
/steering-update
# Compare
diff /tmp/arch-before.md .claude/steering/ARCHITECTURE.md
```
## Performance Metrics
Typical incremental update performance:
| Changes | Time | Tokens | vs Full |
|---------|------|--------|---------|
| 1-5 files | 3-5 min | 5K | 90% faster |
| 6-15 files | 5-8 min | 15K | 82% faster |
| 16-30 files | 8-12 min | 25K | 73% faster |
| 31-50 files | 12-20 min | 40K | 56% faster |
| 50+ files | Consider full regeneration | - | - |
---
**Keep your context fresh:** Run `/steering-update` regularly!