Is it 1 file, trivial, zero risk, no validation needed?
├─ YES → Load atlas-quick skill
└─ NO → Continue...

Does it need validation but not research/planning?
├─ YES → Load atlas-iterative skill
└─ NO → Continue...

Is it 2-5 files with clear requirements?
├─ YES → Load atlas-standard skill ⭐ DEFAULT
└─ NO → Continue...

Is it 6+ files, security-critical, or needs formal requirements?
├─ YES → Load atlas-full skill
└─ NOT SURE → Load atlas-standard skill

User: "Fix the bug where sync fails with empty data"
→ You analyze: 2-5 files, needs research, bug fix
→ You invoke: atlas-standard skill
→ You execute: 5-phase Standard workflow

User: "Fix typo in login button. Use Atlas Quick workflow."
→ You invoke: atlas-quick skill
→ You execute: 2-phase Quick workflow

# Project Atlas Conventions

## Code Standards
- Field naming: [Your conventions]
- State management: [Your patterns]
- Error handling: [Your approach]

## Platform-Specific Rules
- iOS: [Platform gotchas]
- Android: [Platform gotchas]
- Web: [Platform gotchas]

## Deployment Process
- Pre-deployment: [Your checklist]
- Deployment command: [Your command]
- Post-deployment: [Your verification]

## Quality Gates
- Linting: [Your command]
- Type checking: [Your command]
- Testing: [Your command]
- Coverage requirements: [Your threshold]

## Design Standards
- Accessibility: [Your rules]
- Color scheme: [Your palette]
- Typography: [Your fonts]

## Critical Patterns
- Authentication: [Your approach]
- Data persistence: [Your strategy]
- API integration: [Your patterns]

# Project Atlas Conventions

See project documentation:
- Code standards: `/docs/coding-standards.md`
- Architecture: `/docs/architecture.md`
- Deployment: `/docs/deployment.md`
- Testing: `/docs/testing.md`

"Escalating to [TIER] workflow. [REASON: scope expanded, security implications, etc.]"

User: "The sync system loses data during conflict resolution"

Your analysis:
- Scope: Affects sync service, maybe conflict logic (2-5 files)
- Complexity: Needs research to understand conflict handling
- Risk: Medium (data integrity issue)
- Validation: Needs peer review
→ Decision: atlas-standard

Your response:
"I'll use the Atlas Standard workflow for this bug fix."
[Invoke atlas-standard skill]
[Execute 5-phase workflow]

User: "Change primary button color to #007AFF. Use Atlas Quick."

Your analysis:
- Explicit tier specified: Quick
- Verification: 1 file, trivial, zero risk ✅
→ Decision: atlas-quick

Your response:
"Using Atlas Quick workflow for this trivial change."
[Invoke atlas-quick skill]
[Execute 2-phase workflow]

User: "Fix the modal padding issue"

Your initial analysis:
- Expected: 1 file, simple CSS change
→ Initial decision: atlas-quick

During implementation:
- Found: Affects 4 modal components + platform-specific styles
- Complexity: Different platforms use different flex rules
- Risk: Could break modal layouts across platforms
→ Escalation: atlas-standard

Your response:
"Escalating to Standard workflow. Found 4 files affected with platform-specific considerations."
[Invoke atlas-standard skill]
[Restart from Phase 1: Research]

User: "Update user profile state management"

Your first action:
1. Check if `.atlas/conventions.md` exists
2. If yes: Read state management conventions
3. If no: Ask user about state management approach

Your response:
"I'll update the user profile state. First, let me check your project's state management conventions..."
[Check for .atlas/conventions.md]
[Apply project-specific patterns]

"[Describe your task]"
→ Orchestrator selects tier and executes workflow

OR

"[Task description]. Use Atlas [Quick|Iterative|Standard|Full] workflow."
→ Orchestrator uses specified tier