❌ BAD: "Can you describe what analyze_data.py does?"
    Result: Long description, no actionable output

✅ GOOD: "Prepare input file for data analysis and provide command"
    Result: Working configuration + executable command + actual results

1. ⭐⭐⭐⭐⭐ AI prepares input YAML file
   └─ Following template in templates/input_config.yaml
   └─ Validated against schema
   └─ Version controlled in config/input/

2. ⭐⭐⭐⭐⭐ AI provides exact bash command
   └─ Points to correct script in scripts/
   └─ References prepared input file
   └─ Includes all necessary flags

3. ⭐⭐⭐⭐⭐ User executes command
   └─ Copy/paste provided command
   └─ Review output and results
   └─ Version control any changes

4. ⭐⭐⭐⭐⭐ Use Claude for git operations
   └─ Commit results
   └─ Create meaningful commit messages
   └─ Manage branches and PRs

## Task Context
- Repository: digitalmodel (Work)
- Complexity: Medium
- Time sensitivity: Production hotfix
- Dependencies: None
- Testing required: Yes

## Specifications
[Full specifications here]

## Output Format
[Exact format needed]

## Constraints
[Any limitations]

Generate [specific deliverable] following this context.

I need to perform the following operations across multiple repositories:

## Scope
- Repositories: [list or "all work" or "all personal"]
- Operation type: [commit/sync/test/build/deploy]

## Configuration
```yaml
operation: batch_commit
scope: work_repositories
config:
  message: "Update dependencies to latest"
  auto_push: true
  run_tests: true

## Anti-Patterns to Avoid

### ❌ Description-Only Requests

### ❌ Skipping Questions

### ❌ Making Assumptions

## Usage Monitoring Commands

```bash
# Check usage
./scripts/monitoring/check_claude_usage.sh check

# View today's summary
./scripts/monitoring/check_claude_usage.sh today

# View recommendations
./scripts/monitoring/check_claude_usage.sh rec

# Log a task
./scripts/monitoring/check_claude_usage.sh log sonnet digitalmodel "Feature work"