Proficiency Levels
Your PRISM score maps to a proficiency tier that reflects your AI collaboration maturity.
The five tiers
Section titled “The five tiers”Novice (0.0–2.9)
Section titled “Novice (0.0–2.9)”Characteristics:
- Vague, multi-task prompts (“fix everything and make it faster”)
- No file paths or function names in prompts
- Retry storms — repeating the same prompt when it fails
- No verification of AI output
- No use of CLAUDE.md, plan mode, or tools
Token impact: 3–5x more tokens than necessary
Practitioner (3.0–4.9)
Section titled “Practitioner (3.0–4.9)”Characteristics:
- Some specificity in prompts but still vague
- Occasionally reviews output before accepting
- Starting to use
/compactfor context management - Limited recovery strategy when things go wrong
Token impact: ~2.5x baseline
Proficient (5.0–6.9)
Section titled “Proficient (5.0–6.9)”Characteristics:
- Prompts include file paths and function names
- Single-task prompts most of the time
- Regular output review and test runs
- Uses plan mode for complex tasks
- Recovers from errors by adding constraints
Token impact: ~1.8x baseline
Expert (7.0–8.9)
Section titled “Expert (7.0–8.9)”Characteristics:
- Highly specific prompts with context, constraints, and expected behavior
- Consistent single-task decomposition
- Verification after every significant change
- Uses subagents and skills for delegation
- Maintains CLAUDE.md with project context
Token impact: ~1.3x baseline
Elite (9.0–10.0)
Section titled “Elite (9.0–10.0)”Characteristics:
- Prompts read like well-written tickets — specific, scoped, testable
- Minimal corrections needed
- Full verification workflow (tests, type-check, manual review)
- Expert use of all available tools and configuration
- Context management is second nature
Token impact: ~1x (baseline efficient usage)
Moving up tiers
Section titled “Moving up tiers”The most impactful improvements by tier:
| From | Focus on | Expected lift |
|---|---|---|
| Novice → Practitioner | Add file paths and function names to prompts | +1.5–2.0 PQ |
| Practitioner → Proficient | One task per prompt, run tests after changes | +1.0 PQ, +1.5 VD |
| Proficient → Expert | Use plan mode, CLAUDE.md, recovery strategies | +1.0 AF, +1.0 IE |
| Expert → Elite | Delegate with subagents, optimize context management | +0.5 AF, +0.5 TU |
The biggest gains come early — moving from Novice to Proficient can cut token spend by 60%.