Every function is tagged with a set of effects based on the built-in functions it calls. The mutation engine uses this classification to skip flaky mutants, seal nondeterminism at test time, and compute an effect stability score that feeds into the integrity grade.

For an overview of all advanced features, see [Advanced overview](/docs/mutation/advanced/).

## What it does

The engine uses fixpoint analysis to propagate effects through the call graph. Each function receives one or more effect tags:

| Effect | Triggered by | Example built-ins |
|--------|-------------|-------------------|
| `PURE` | No side effects | `abs`, `len`, `str`, `sorted`, `range` |
| `IO` | Console/output | `print`, `println`, `input` |
| `TIME` | Clock access | `time`, `sleep` |
| `RANDOM` | Nondeterminism | `rand_int`, `rand_seed` |
| `NETWORK` | HTTP calls | `_http_request` |
| `FS` | File system | `read_text`, `write_text`, `list_dir`, `mkdir` |
| `PROCESS` | OS interaction | `exit`, `getenv`, `argv`, `spawn` |

Effects propagate transitively: a function that calls a `RANDOM` function is also tagged `RANDOM`.

## What it outputs

### Nondeterministic mutant filtering

Mutants inside functions tagged `TIME`, `RANDOM`, or `NETWORK` are skipped by default because they produce flaky results:

```text
Skipped 3 nondeterministic mutants (use --unstable to include)
```

### Runtime sealing

For `RANDOM`-effect functions, the engine sets a deterministic seed before running each mutant. This makes `rand_int()` repeatable and `rand_seed()` a no-op during mutation testing, so `RANDOM` mutants produce stable results even without `--unstable`.

### Effect stability score

The effect stability metric measures the fraction of functions with stable (non-flaky) effects. It contributes 20% to the integrity score.

### Integrity scoring

The integrity score is a weighted composite of four metrics:

| Component | Weight | Source |
|-----------|--------|--------|
| Mutation score | 40% | Fraction of mutants killed |
| Contract adequacy | 20% | Fraction of contract obligations defended |
| Oracle coverage | 20% | Fraction of tests with strong oracles |
| Effect stability | 20% | Fraction of functions with stable effects |

The score produces a letter grade:

| Grade | Score range |
|-------|------------|
| A | 90 to 100 |
| B | 75 to 89 |
| C | 60 to 74 |
| D | 40 to 59 |
| F | 0 to 39 |

The integrity score is included in mutation certificates and the manifest. Set a minimum:

```toml
[mutation]
min_integrity = 75.0   # Fail if integrity score < 75 (grade B)
```

When the threshold is set, `vary mutate` exits with code 1 if the integrity score falls below it. The `--why` output includes a component breakdown showing which area is weakest.

## How to use it

Effect classification runs automatically. To include nondeterministic mutants:

```bash
vary mutate source.vary --tests test.vary --unstable
```

To see which functions have which effects, use `--observe`:

```bash
vary mutate source.vary --tests test.vary --observe
```

## How to interpret the results

| You see... | What to do |
|------------|-----------|
| Mutants skipped as nondeterministic | These functions use TIME/RANDOM/NETWORK; use `--unstable` to include them or accept the filtering |
| Low effect stability | Many functions use nondeterministic built-ins; consider isolating side effects behind interfaces |
| `RANDOM` mutants still flaky with sealing | The function may use external randomness not captured by `rand_seed`; quarantine or isolate |
| Integrity grade below threshold | Check the component breakdown: the weakest component tells you where to focus |

## What to do next

| Page | Topic |
|------|-------|
| [Signatures](/docs/mutation/signatures/) | How mutant identities survive refactors |
| [Contracts](/docs/mutation/contracts/) | Contract adequacy contributes 20% to the integrity score |
| [Oracle analysis](/docs/mutation/oracle-analysis/) | Oracle coverage contributes 20% to the integrity score |