---
name: code-quality-crap
description: >-
  CRAP score quality gate for code complexity and test coverage. Use when
  reviewing methods for complexity risk, deciding whether to split methods,
  or interpreting complexity and coverage risk without manufacturing tests.
---

# Code Quality: CRAP Score

CRAP (Change Risk Anti-Patterns) measures the risk of a method based on
cyclomatic complexity and test coverage.

## When to Use

- Reviewing new or changed methods for complexity risk
- Deciding whether a method needs splitting
- Interpreting coverage risk for a given complexity
- During `PURE_REFACTOR` implementation or candidate review

## Formula

$$\text{CRAP}(m) = \text{comp}(m)^2 \times (1 - \text{cov}(m))^3 + \text{comp}(m)$$

Where:

- $\text{comp}(m)$ = cyclomatic complexity of method $m$
- $\text{cov}(m)$ = test coverage ratio (0 to 1) of method $m$

## Quality Gate

Every method MUST have a CRAP score below 6.

## Coverage Thresholds

These thresholds are intentionally stricter than the mathematical minimum,
providing a safety margin:

| Complexity | Recommended coverage for CRAP < 6 |
|-----------|-----------------------------------|
| 1 | 0% (any) |
| 2 | > 0% |
| 3 | 40%+ |
| 4 | 60%+ |
| 5 | 80%+ |
| 6+ | **Not achievable**: must split to reduce complexity |

## Actions by Scenario

| Scenario | Action |
|----------|--------|
| CRAP >= 6 due to high complexity | Extract complex branches into smaller methods |
| CRAP >= 6 due to low coverage | First reduce complexity; if important behavior lacks evidence, route a separate verification contract rather than adding coverage-only tests |
| Complexity 6+ regardless of coverage | Must split: CRAP floor equals complexity |

## Severity Classification

- **Medium**: complexity 4–5 without proportional test coverage
- **High**: complexity 6+ (must split, no amount of testing fixes it)