Testing with Attesta

Attesta is designed to be testable from day one. Every component — scorers, renderers, challenges, and the audit logger — uses protocol-based interfaces that are trivial to mock. This guide covers testing patterns from unit tests through integration tests.

Auto-Approve in Tests

For Python, the simplest approach is relying on no-TTY behavior: in CI environments, stdin is not a TTY, so Attesta automatically uses _DefaultRenderer (auto-approve).

from attesta import gate

@gate
def deploy(service: str) -> str:
    return f"Deployed {service}"

# In CI (no TTY, Python): auto-approved, the function runs normally
result = deploy("api-gateway")
assert result == "Deployed api-gateway"

In TypeScript, no-TTY behavior defaults to deny. For CI tests in TypeScript, provide a test renderer that approves or denies deterministically.

Mock Renderers

For explicit control over approval outcomes, use mock renderers.

ApproveAllRenderer

A renderer that approves every action, regardless of risk level:

from attesta import (
    ActionContext, RiskAssessment, ChallengeType,
    ChallengeResult, Verdict,
)

class ApproveAllRenderer:
    """Approves all actions. Use in tests to bypass human interaction."""

    async def render_approval(
        self, ctx: ActionContext, risk: RiskAssessment
    ) -> Verdict:
        return Verdict.APPROVED

    async def render_challenge(
        self,
        ctx: ActionContext,
        risk: RiskAssessment,
        challenge_type: ChallengeType,
    ) -> ChallengeResult:
        return ChallengeResult(
            passed=True,
            challenge_type=challenge_type,
            responder="test-auto",
        )

    async def render_info(self, message: str) -> None:
        pass

    async def render_auto_approved(
        self, ctx: ActionContext, risk: RiskAssessment
    ) -> None:
        pass

DenyAllRenderer

A renderer that denies every action:

class DenyAllRenderer:
    """Denies all actions. Use to test denial handling."""

    async def render_approval(
        self, ctx: ActionContext, risk: RiskAssessment
    ) -> Verdict:
        return Verdict.DENIED

    async def render_challenge(
        self,
        ctx: ActionContext,
        risk: RiskAssessment,
        challenge_type: ChallengeType,
    ) -> ChallengeResult:
        return ChallengeResult(
            passed=False,
            challenge_type=challenge_type,
            responder="test-deny",
        )

    async def render_info(self, message: str) -> None:
        pass

    async def render_auto_approved(
        self, ctx: ActionContext, risk: RiskAssessment
    ) -> None:
        pass

RecordingAuditLogger

An audit logger that records entries in memory for assertions:

from attesta import ActionContext, ApprovalResult

class RecordingAuditLogger:
    """Records all audit entries in memory for test assertions."""

    def __init__(self):
        self.entries: list[tuple[ActionContext, ApprovalResult]] = []

    async def log(self, ctx: ActionContext, result: ApprovalResult) -> str:
        self.entries.append((ctx, result))
        return f"test-audit-{len(self.entries)}"

Testing Gated Functions

Test That Approved Actions Execute

import pytest
from attesta import gate, AttestaDenied

@gate(renderer=ApproveAllRenderer())
def transfer_funds(from_account: str, to_account: str, amount: float) -> str:
    return f"Transferred ${amount} from {from_account} to {to_account}"

def test_approved_transfer():
    result = transfer_funds("acc-1", "acc-2", 100.0)
    assert result == "Transferred $100.0 from acc-1 to acc-2"
    assert "acc-1" in result

Test That Denied Actions Raise

@gate(renderer=DenyAllRenderer(), risk="high")
def delete_database(db_name: str) -> str:
    return f"Deleted {db_name}"

def test_denied_raises_exception():
    with pytest.raises(AttestaDenied) as exc_info:
        delete_database("production")

    assert exc_info.value.result is not None
    assert exc_info.value.result.verdict.value == "denied"

Test Risk Level Assignment

@gate(
    renderer=ApproveAllRenderer(),
    audit_logger=RecordingAuditLogger(),
    risk="critical",
)
async def nuke_everything() -> str:
    return "done"

@pytest.mark.asyncio
async def test_critical_risk_override():
    audit = nuke_everything.__gate__._audit  # access internal audit logger
    # ... or pass the audit logger explicitly:

    audit = RecordingAuditLogger()

    @gate(
        renderer=ApproveAllRenderer(),
        audit_logger=audit,
        risk="critical",
    )
    async def dangerous_action():
        return "executed"

    result = await dangerous_action()
    assert result == "executed"

    ctx, approval = audit.entries[0]
    assert approval.risk_assessment.level.value == "critical"
    assert approval.risk_assessment.score == 0.90

Test with Async Functions

@pytest.mark.asyncio
async def test_async_gated_function():
    @gate(renderer=ApproveAllRenderer())
    async def fetch_data(url: str) -> str:
        return f"Data from {url}"

    result = await fetch_data("https://api.example.com/users")
    assert "api.example.com" in result

@pytest.mark.asyncio
async def test_async_denied():
    @gate(renderer=DenyAllRenderer(), risk="high")
    async def dangerous_fetch(url: str) -> str:
        return f"Data from {url}"

    with pytest.raises(AttestaDenied):
        await dangerous_fetch("https://api.example.com/admin")

Testing Custom Scorers

Unit Testing a Scorer

from attesta import ActionContext, RiskScorer

def test_scorer_protocol_compliance():
    """Every custom scorer must satisfy the RiskScorer protocol."""
    scorer = MyCustomScorer()
    assert isinstance(scorer, RiskScorer)
    assert isinstance(scorer.name, str)
    assert len(scorer.name) > 0

def test_scorer_returns_bounded_values():
    """Scores must always be in [0.0, 1.0]."""
    scorer = MyCustomScorer()
    test_cases = [
        ActionContext(function_name="get_user"),
        ActionContext(function_name="delete_database", environment="production"),
        ActionContext(
            function_name="transfer",
            hints={"amount": 1_000_000, "pii": True},
        ),
        ActionContext(function_name=""),  # edge case: empty name
    ]

    for ctx in test_cases:
        score = scorer.score(ctx)
        assert 0.0 <= score <= 1.0, (
            f"Score {score} out of bounds for {ctx.function_name}"
        )

def test_scorer_risk_ordering():
    """Destructive actions should score higher than read-only actions."""
    scorer = MyCustomScorer()
    read_ctx = ActionContext(function_name="get_user")
    write_ctx = ActionContext(function_name="update_user")
    delete_ctx = ActionContext(function_name="delete_user")

    read_score = scorer.score(read_ctx)
    write_score = scorer.score(write_ctx)
    delete_score = scorer.score(delete_ctx)

    assert read_score < write_score
    assert write_score < delete_score

Testing DefaultRiskScorer Factors

from attesta.core.risk import DefaultRiskScorer

def test_default_scorer_novelty_decreases():
    """Novelty score should decrease with repeated calls."""
    scorer = DefaultRiskScorer()
    ctx = ActionContext(function_name="deploy")

    first_score = scorer.score(ctx)
    second_score = scorer.score(ctx)
    tenth_score = scorer.score(ctx)

    # Novelty decreases, so total score should decrease
    # (assuming other factors remain constant)
    assert first_score >= second_score

def test_default_scorer_reset_novelty():
    """reset_novelty() should clear call counters."""
    scorer = DefaultRiskScorer()
    ctx = ActionContext(function_name="deploy")

    scorer.score(ctx)
    scorer.score(ctx)
    scorer.reset_novelty()

    # After reset, novelty should be high again
    assessment = scorer.assess(ctx)
    novelty_factor = next(
        f for f in assessment.factors if f.name == "novelty"
    )
    assert "0 time(s)" in novelty_factor.evidence

Testing Composite Scorers

from attesta.core.risk import (
    CompositeRiskScorer, FixedRiskScorer, MaxRiskScorer,
)

def test_composite_weighted_average():
    """CompositeRiskScorer should produce a weighted average."""
    scorer = CompositeRiskScorer(
        scorers=[
            (FixedRiskScorer(0.2), 1.0),
            (FixedRiskScorer(0.8), 1.0),
        ]
    )
    ctx = ActionContext(function_name="test")
    score = scorer.score(ctx)
    assert abs(score - 0.5) < 0.01  # (0.2 + 0.8) / 2

def test_max_scorer_takes_highest():
    """MaxRiskScorer should return the maximum child score."""
    scorer = MaxRiskScorer(
        scorers=[FixedRiskScorer(0.3), FixedRiskScorer(0.7)]
    )
    ctx = ActionContext(function_name="test")
    assert scorer.score(ctx) == 0.7

Testing Custom Challenges

Mock the I/O Layer

Challenges that use input() need mocking in tests:

import pytest
from attesta import (
    ActionContext, RiskAssessment, RiskLevel, ChallengeProtocol,
)

@pytest.mark.asyncio
async def test_challenge_with_mocked_input(monkeypatch):
    """Test a challenge by mocking user input."""
    from attesta.challenges.confirm import ConfirmChallenge

    challenge = ConfirmChallenge(min_review_seconds=0.0)

    # Mock input to return "y"
    monkeypatch.setattr("builtins.input", lambda _: "y")

    ctx = ActionContext(function_name="deploy")
    risk = RiskAssessment(score=0.45, level=RiskLevel.MEDIUM)

    result = await challenge.present(ctx, risk)
    assert result.passed is True
    assert result.challenge_type.value == "confirm"

@pytest.mark.asyncio
async def test_challenge_denial(monkeypatch):
    """Test that 'n' input results in denial."""
    from attesta.challenges.confirm import ConfirmChallenge

    challenge = ConfirmChallenge(min_review_seconds=0.0)
    monkeypatch.setattr("builtins.input", lambda _: "n")

    ctx = ActionContext(function_name="deploy")
    risk = RiskAssessment(score=0.45, level=RiskLevel.MEDIUM)

    result = await challenge.present(ctx, risk)
    assert result.passed is False

Test ChallengeResult Details

@pytest.mark.asyncio
async def test_challenge_records_response_time(monkeypatch):
    from attesta.challenges.confirm import ConfirmChallenge

    challenge = ConfirmChallenge(min_review_seconds=0.0)
    monkeypatch.setattr("builtins.input", lambda _: "y")

    ctx = ActionContext(function_name="deploy")
    risk = RiskAssessment(score=0.45, level=RiskLevel.MEDIUM)

    result = await challenge.present(ctx, risk)
    assert result.response_time_seconds >= 0.0
    assert result.questions_asked >= 0

Testing the Full Pipeline

Integration Test with evaluate()

import pytest
from attesta.core.gate import Attesta
from attesta import ActionContext, Verdict, RiskLevel

@pytest.mark.asyncio
async def test_full_pipeline_low_risk():
    """Low risk should auto-approve without challenge."""
    audit = RecordingAuditLogger()
    attesta = Attesta(
        renderer=ApproveAllRenderer(),
        audit_logger=audit,
    )

    ctx = ActionContext(function_name="get_user")
    result = await attesta.evaluate(ctx)

    assert result.verdict == Verdict.APPROVED
    assert result.challenge_result is None  # auto-approved
    assert result.audit_entry_id is not None
    assert len(audit.entries) == 1

@pytest.mark.asyncio
async def test_full_pipeline_high_risk():
    """High risk should go through a QUIZ challenge."""
    audit = RecordingAuditLogger()
    attesta = Attesta(
        renderer=ApproveAllRenderer(),
        audit_logger=audit,
        risk_override=RiskLevel.HIGH,
    )

    ctx = ActionContext(function_name="deploy_to_production")
    result = await attesta.evaluate(ctx)

    assert result.verdict == Verdict.APPROVED
    assert result.challenge_result is not None
    assert result.challenge_result.challenge_type.value == "quiz"

@pytest.mark.asyncio
async def test_full_pipeline_denied():
    """Denied actions should have DENIED verdict and no execution."""
    attesta = Attesta(
        renderer=DenyAllRenderer(),
        risk_override=RiskLevel.HIGH,
    )

    ctx = ActionContext(function_name="delete_users")
    result = await attesta.evaluate(ctx)

    assert result.verdict == Verdict.DENIED
    assert result.challenge_result is not None
    assert result.challenge_result.passed is False

Integration Test with Trust Engine

import pytest
from attesta.core.gate import Attesta
from attesta.core.trust import TrustEngine
from attesta import ActionContext, RiskLevel

@pytest.mark.asyncio
async def test_trust_reduces_risk():
    """A trusted agent should get lower effective risk."""
    engine = TrustEngine(initial_score=0.3, influence=0.3)

    # Build up trust with successful actions
    for _ in range(20):
        engine.record_success(
            agent_id="trusted-bot",
            action_name="safe_action",
            domain="general",
            risk_score=0.3,
        )

    attesta = Attesta(
        renderer=ApproveAllRenderer(),
        trust_engine=engine,
    )

    ctx = ActionContext(
        function_name="moderate_action",
        agent_id="trusted-bot",
        environment="production",
    )
    result = await attesta.evaluate(ctx)

    # With high trust, the effective risk should be reduced
    # (exact value depends on the scorer, but trust should lower it)
    assert result.verdict == Verdict.APPROVED

@pytest.mark.asyncio
async def test_trust_never_downgrades_critical():
    """CRITICAL risk must never be downgraded by trust."""
    engine = TrustEngine(initial_score=0.9, influence=0.5)

    attesta = Attesta(
        renderer=ApproveAllRenderer(),
        trust_engine=engine,
        risk_override=RiskLevel.CRITICAL,
    )

    ctx = ActionContext(
        function_name="delete_everything",
        agent_id="super-trusted-bot",
    )
    result = await attesta.evaluate(ctx)

    # Even with max trust, CRITICAL should remain CRITICAL
    assert result.risk_assessment.level == RiskLevel.CRITICAL

Testing with from_config()

Test that your YAML configuration loads correctly:

import tempfile
import pytest
from pathlib import Path
from attesta import Attesta

def test_from_config_loads_yaml():
    config = """
policy:
  fail_mode: deny
  minimum_review_seconds:
    medium: 3
    high: 10
    critical: 30

trust:
  initial_score: 0.3
  ceiling: 0.85
  influence: 0.25
"""
    with tempfile.NamedTemporaryFile(
        mode="w", suffix=".yaml", delete=False
    ) as f:
        f.write(config)
        f.flush()
        attesta = Attesta.from_config(f.name)

    assert attesta.policy.get("policy", {}).get("fail_mode") == "deny"

def test_from_config_missing_file():
    with pytest.raises(FileNotFoundError):
        Attesta.from_config("/nonexistent/attesta.yaml")

Pytest Configuration

Add these fixtures to your conftest.py for reuse across tests:

conftest.py

import pytest
from attesta import (
    ActionContext, RiskAssessment, ApprovalResult,
    ChallengeResult, ChallengeType, RiskLevel, Verdict,
)

@pytest.fixture
def approve_renderer():
    """Renderer that auto-approves everything."""
    return ApproveAllRenderer()

@pytest.fixture
def deny_renderer():
    """Renderer that denies everything."""
    return DenyAllRenderer()

@pytest.fixture
def recording_audit():
    """Audit logger that records entries in memory."""
    return RecordingAuditLogger()

@pytest.fixture
def sample_context():
    """A sample ActionContext for testing."""
    return ActionContext(
        function_name="deploy",
        args=("api-gateway", "2.0.0"),
        function_doc="Deploy a service to production.",
        environment="production",
        agent_id="test-agent",
        hints={"production": True},
    )

@pytest.fixture
def low_risk():
    return RiskAssessment(score=0.15, level=RiskLevel.LOW)

@pytest.fixture
def high_risk():
    return RiskAssessment(score=0.75, level=RiskLevel.HIGH)

@pytest.fixture
def critical_risk():
    return RiskAssessment(score=0.92, level=RiskLevel.CRITICAL)

Put ApproveAllRenderer, DenyAllRenderer, and RecordingAuditLogger in a shared test utilities module. They are useful across your entire test suite, not just Attesta-specific tests.

Summary

Testing Scenario	Approach
CI/CD auto-approval	No setup needed — non-TTY environments auto-approve
Test approved execution	`ApproveAllRenderer`
Test denial handling	`DenyAllRenderer` + `pytest.raises(AttestaDenied)`
Verify audit records	`RecordingAuditLogger` + assert on `.entries`
Test risk levels	Use `risk="high"` override + check `result.risk_assessment.level`
Test custom scorers	Unit test with crafted `ActionContext` instances
Test custom challenges	`monkeypatch` input, check `ChallengeResult`
Full pipeline	`Attesta.evaluate()` with mock renderer + audit logger
Trust engine	Pre-populate trust history, verify risk adjustment

Protocols

Protocol definitions for mock implementation

Production Deployment

Deploy with confidence after thorough testing

Getting Started

Core Concepts

Configuration

Domain Profiles

Examples

Guides

TypeScript

Resources

Testing with Attesta

Auto-Approve in Tests

Mock Renderers

ApproveAllRenderer

DenyAllRenderer

RecordingAuditLogger

Testing Gated Functions

Test That Approved Actions Execute

Test That Denied Actions Raise

Test Risk Level Assignment

Test with Async Functions

Testing Custom Scorers

Unit Testing a Scorer

Testing DefaultRiskScorer Factors

Testing Composite Scorers

Testing Custom Challenges

Mock the I/O Layer

Test ChallengeResult Details

Testing the Full Pipeline

Integration Test with evaluate()

Integration Test with Trust Engine

Testing with from_config()

Pytest Configuration

Summary

Protocols

Production Deployment

Getting Started

Core Concepts

Configuration

Domain Profiles

Examples

Guides

TypeScript

Resources

​Auto-Approve in Tests

​Mock Renderers

​ApproveAllRenderer

​DenyAllRenderer

​RecordingAuditLogger

​Testing Gated Functions

​Test That Approved Actions Execute

​Test That Denied Actions Raise

​Test Risk Level Assignment

​Test with Async Functions

​Testing Custom Scorers

​Unit Testing a Scorer

​Testing DefaultRiskScorer Factors

​Testing Composite Scorers

​Testing Custom Challenges

​Mock the I/O Layer

​Test ChallengeResult Details

​Testing the Full Pipeline

​Integration Test with evaluate()

​Integration Test with Trust Engine

​Testing with from_config()

​Pytest Configuration

​Summary

Protocols

Production Deployment

Auto-Approve in Tests

Mock Renderers

ApproveAllRenderer

DenyAllRenderer

RecordingAuditLogger

Testing Gated Functions

Test That Approved Actions Execute

Test That Denied Actions Raise

Test Risk Level Assignment

Test with Async Functions

Testing Custom Scorers

Unit Testing a Scorer

Testing DefaultRiskScorer Factors

Testing Composite Scorers

Testing Custom Challenges

Mock the I/O Layer

Test ChallengeResult Details

Testing the Full Pipeline

Integration Test with evaluate()

Integration Test with Trust Engine

Testing with from_config()

Pytest Configuration

Summary