feat: plane-2 tool-call intent validation (path containment slice)

Implements the full vertical slice for invariant #9 tool-call assessment:
- core:toolintent — new module with ToolCallRule seam, ToolCallAssessor,
  WorldProbe/FileSystemWorldProbe, WorkspacePolicy, PathContainmentRule,
  and RiskMapping (assessment → RiskSummary / AssessedIssue)
- core:tools — ToolCallAssessmentRecord + ToolInvocationRecord.assessment
  field; DefaultToolReducer handles ToolCallAssessedEvent (replay proof)
- core:config — ToolsConfig gains workspaceRoot + privilegedLocations;
  ConfigLoader parses both; DEFAULT_PRIVILEGED_LOCATIONS built-in
- core:kernel — OrchestratorEngines gains toolCallAssessor/workspacePolicy/
  worldProbe fields; SessionOrchestrator.dispatchToolCalls runs
  runPlane2Assessment before the tier gate: BLOCK → hard-reject without
  executing; PROMPT_USER → elevates tier into approval path with plane2Risk
  in the ApprovalRequestedEvent
- apps/server — constructs PathContainmentRule + ToolCallAssessor +
  WorkspacePolicy from config and wires them into OrchestratorEngines

Assessment is recorded as ToolCallAssessedEvent (environment observed once,
facts stored, replay reads events — invariant #9). Assessor and WorldProbe
are only invoked on the live orchestrator path, never in replay.
This commit is contained in:
2026-05-31 04:22:58 +04:00
parent 2964849bb2
commit 545068d222
24 changed files with 931 additions and 3 deletions
+3
View File
@@ -14,7 +14,10 @@ dependencies {
testImplementation(project(":core:inference"))
testImplementation(project(":core:kernel"))
testImplementation(project(":core:risk"))
testImplementation(project(":core:tools"))
testImplementation(project(":core:toolintent"))
testImplementation(project(":infrastructure:persistence"))
testImplementation("org.jetbrains.kotlin:kotlin-test")
testImplementation(project(":core:artifacts"))
testImplementation(project(":core:artifacts-store"))
testImplementation(project(":testing:fixtures"))
@@ -0,0 +1,277 @@
import com.correx.core.approvals.ApprovalProjector
import com.correx.core.approvals.DefaultApprovalReducer
import com.correx.core.approvals.DefaultApprovalRepository
import com.correx.core.approvals.Tier
import com.correx.core.approvals.domain.DefaultApprovalEngine
import com.correx.core.artifacts.DefaultArtifactReducer
import com.correx.core.events.events.ApprovalRequestedEvent
import com.correx.core.events.events.ToolCallAssessedEvent
import com.correx.core.events.events.ToolExecutionRejectedEvent
import com.correx.core.events.events.ToolRequest
import com.correx.core.events.execution.RetryPolicy
import com.correx.core.events.risk.RiskAction
import com.correx.core.events.types.ProviderId
import com.correx.core.events.types.SessionId
import com.correx.core.events.types.StageId
import com.correx.core.events.types.TransitionId
import com.correx.core.inference.CapabilityScore
import com.correx.core.inference.FinishReason
import com.correx.core.inference.InferenceProvider
import com.correx.core.inference.InferenceRepository
import com.correx.core.inference.InferenceRequest
import com.correx.core.inference.InferenceResponse
import com.correx.core.inference.InferenceRouter
import com.correx.core.inference.InferenceState
import com.correx.core.inference.ModelCapability
import com.correx.core.inference.ProviderHealth
import com.correx.core.inference.Tokenizer
import com.correx.core.inference.ToolCallFunction
import com.correx.core.inference.ToolCallRequest
import com.correx.core.inference.TokenUsage
import com.correx.core.kernel.orchestration.DefaultOrchestrationReducer
import com.correx.core.kernel.orchestration.DefaultSessionOrchestrator
import com.correx.core.kernel.orchestration.OrchestrationConfig
import com.correx.core.kernel.orchestration.OrchestrationProjector
import com.correx.core.kernel.orchestration.OrchestrationRepository
import com.correx.core.kernel.orchestration.OrchestratorEngines
import com.correx.core.kernel.orchestration.OrchestratorRepositories
import com.correx.core.kernel.retry.DefaultRetryCoordinator
import com.correx.core.risk.DefaultRiskAssessor
import com.correx.core.sessions.DefaultSessionReducer
import com.correx.core.sessions.DefaultSessionRepository
import com.correx.core.sessions.SessionProjector
import com.correx.core.sessions.projections.replay.DefaultEventReplayer
import com.correx.core.sessions.projections.replay.EventReplayer
import com.correx.core.toolintent.ToolCallAssessment
import com.correx.core.toolintent.ToolCallAssessmentInput
import com.correx.core.toolintent.ToolCallAssessor
import com.correx.core.toolintent.ToolCallRule
import com.correx.core.toolintent.WorkspacePolicy
import com.correx.core.tools.contract.Tool
import com.correx.core.tools.contract.ToolCapability
import com.correx.core.tools.contract.ToolExecutor
import com.correx.core.tools.contract.ToolResult
import com.correx.core.tools.contract.ValidationResult
import com.correx.core.tools.registry.ToolRegistry
import com.correx.core.transitions.graph.StageConfig
import com.correx.core.transitions.graph.TransitionEdge
import com.correx.core.transitions.graph.WorkflowGraph
import com.correx.core.transitions.resolution.DefaultTransitionResolver
import com.correx.core.validation.pipeline.ValidationPipeline
import com.correx.infrastructure.persistence.InMemoryEventStore
import com.correx.infrastructure.persistence.artifact.LiveArtifactRepository
import com.correx.testing.contracts.fixtures.artifactstore.NoopArtifactStore
import com.correx.testing.fixtures.context.ContextFixtures
import com.correx.testing.fixtures.cyclePolicyMissingValidator
import com.correx.testing.fixtures.inference.MockTokenizer
import kotlinx.coroutines.launch
import kotlinx.coroutines.runBlocking
import kotlinx.coroutines.withTimeout
import kotlinx.coroutines.yield
import kotlinx.serialization.json.JsonObject
import kotlinx.serialization.json.buildJsonObject
import java.nio.file.Path
import java.util.concurrent.atomic.AtomicBoolean
import kotlin.test.Test
import kotlin.test.assertEquals
import kotlin.test.assertNotNull
import kotlin.test.assertNull
import kotlin.test.assertTrue
@Suppress("LongMethod")
class ToolCallGateTest {
private inner class FakeFileWriteTool(override val tier: Tier = Tier.T1) : Tool {
override val name = "file_write"
override val description = "fake file write"
override val parametersSchema: JsonObject = buildJsonObject {}
override val requiredCapabilities: Set<ToolCapability> = setOf(ToolCapability.FILE_WRITE)
override fun validateRequest(request: ToolRequest): ValidationResult = ValidationResult.Valid
}
private inner class FakeToolRegistry(private val tool: Tool) : ToolRegistry {
override fun resolve(name: String): Tool? = if (name == tool.name) tool else null
override fun all(): List<Tool> = listOf(tool)
}
private inner class RecordingExecutor : ToolExecutor {
val executeCalled = AtomicBoolean(false)
override suspend fun execute(request: ToolRequest): ToolResult {
executeCalled.set(true)
return ToolResult.Success(invocationId = request.invocationId, output = "ok", exitCode = 0)
}
}
private inner class ToolCallingProvider(private val toolName: String) : InferenceProvider {
override val id = ProviderId("tool-caller")
override val name = "tool-caller"
override val tokenizer: Tokenizer = MockTokenizer()
private var callCount = 0
override suspend fun infer(request: InferenceRequest): InferenceResponse {
callCount++
return if (callCount == 1) {
InferenceResponse(
requestId = request.requestId,
text = "",
finishReason = FinishReason.ToolCall,
tokensUsed = TokenUsage(10, 5),
latencyMs = 0,
toolCalls = listOf(
ToolCallRequest(
id = "tc-1",
function = ToolCallFunction(name = toolName, arguments = "{}"),
),
),
)
} else {
InferenceResponse(
requestId = request.requestId,
text = "done",
finishReason = FinishReason.Stop,
tokensUsed = TokenUsage(10, 5),
latencyMs = 0,
)
}
}
override suspend fun healthCheck(): ProviderHealth = ProviderHealth.Healthy
override fun capabilities(): Set<CapabilityScore> = setOf(CapabilityScore(ModelCapability.General, 1.0))
}
private fun ruleReturning(action: RiskAction): ToolCallRule = object : ToolCallRule {
override fun appliesTo(capabilities: Set<ToolCapability>) = true
override fun assess(input: ToolCallAssessmentInput) = ToolCallAssessment(disposition = action)
}
private val workspace = Path.of("/work")
private fun buildOrchestrator(
executor: ToolExecutor,
tool: Tool,
assessorRule: ToolCallRule?,
): Pair<DefaultSessionOrchestrator, InMemoryEventStore> {
val eventStore = InMemoryEventStore()
val artifactStore = NoopArtifactStore()
val toolRegistry = FakeToolRegistry(tool)
val assessor = assessorRule?.let { ToolCallAssessor(listOf(it)) }
val policy = if (assessor != null) WorkspacePolicy(workspace) else null
val provider = ToolCallingProvider(tool.name)
val inferenceRouter = object : InferenceRouter {
override suspend fun route(stageId: StageId, requiredCapabilities: Set<ModelCapability>): InferenceProvider =
provider
}
val approvalRepository = DefaultApprovalRepository(
DefaultEventReplayer(eventStore, ApprovalProjector(DefaultApprovalReducer())),
)
val repositories = OrchestratorRepositories(
eventStore = eventStore,
inferenceRepository = InferenceRepository(object : EventReplayer<InferenceState> {
override fun rebuild(sessionId: SessionId) = InferenceState()
}),
orchestrationRepository = OrchestrationRepository(
DefaultEventReplayer(eventStore, OrchestrationProjector(DefaultOrchestrationReducer())),
),
sessionRepository = DefaultSessionRepository(
DefaultEventReplayer(eventStore, SessionProjector(DefaultSessionReducer())),
),
artifactRepository = LiveArtifactRepository(eventStore, DefaultArtifactReducer()),
approvalRepository = approvalRepository,
)
val engines = OrchestratorEngines(
transitionResolver = DefaultTransitionResolver { _, _ -> true },
contextPackBuilder = ContextFixtures.simpleBuilder(),
inferenceRouter = inferenceRouter,
validationPipeline = ValidationPipeline(validators = listOf(cyclePolicyMissingValidator())),
approvalEngine = DefaultApprovalEngine(),
riskAssessor = DefaultRiskAssessor(),
toolExecutor = executor,
toolRegistry = toolRegistry,
toolCallAssessor = assessor,
workspacePolicy = policy,
)
val orchestrator = DefaultSessionOrchestrator(
repositories = repositories,
engines = engines,
retryCoordinator = DefaultRetryCoordinator(eventStore),
artifactStore = artifactStore,
)
return Pair(orchestrator, eventStore)
}
private fun singleStageGraph(allowedTools: Set<String> = setOf("file_write")): WorkflowGraph =
WorkflowGraph(
id = "gate-test",
stages = mapOf(
StageId("A") to StageConfig(allowedTools = allowedTools),
),
transitions = setOf(
TransitionEdge(TransitionId("t1"), StageId("A"), StageId("done"), condition = { true }),
),
start = StageId("A"),
)
@Test
fun `BLOCK rule prevents executor from being called and emits rejected event`(): Unit = runBlocking {
val executor = RecordingExecutor()
val (orchestrator, eventStore) = buildOrchestrator(executor, FakeFileWriteTool(), ruleReturning(RiskAction.BLOCK))
val sessionId = SessionId("gate-block")
val config = OrchestrationConfig(retryPolicy = RetryPolicy(maxAttempts = 1, backoffMs = 0))
orchestrator.run(sessionId, singleStageGraph(), config)
val events = eventStore.read(sessionId)
assertTrue(!executor.executeCalled.get(), "executor must NOT be called when BLOCK")
assertNotNull(events.find { it.payload is ToolCallAssessedEvent }, "ToolCallAssessedEvent must be emitted")
assertNotNull(events.find { it.payload is ToolExecutionRejectedEvent }, "ToolExecutionRejectedEvent must be emitted")
}
@Test
fun `PROMPT_USER rule on T2 tool triggers approval with plane2 risk summary`(): Unit = runBlocking {
val executor = RecordingExecutor()
// Use T2 so the engine does not auto-approve (PROMPT mode auto-approves up to T1 only)
val (orchestrator, eventStore) = buildOrchestrator(
executor, FakeFileWriteTool(Tier.T2), ruleReturning(RiskAction.PROMPT_USER),
)
val sessionId = SessionId("gate-prompt")
val config = OrchestrationConfig(retryPolicy = RetryPolicy(maxAttempts = 1, backoffMs = 0))
// Run in background; orchestrator will pause waiting for approval
val job = launch { orchestrator.run(sessionId, singleStageGraph(), config) }
// Wait for ApprovalRequestedEvent
withTimeout(5_000) {
while (eventStore.read(sessionId).none { it.payload is ApprovalRequestedEvent }) {
yield()
}
}
val approval = eventStore.read(sessionId).firstNotNullOfOrNull { it.payload as? ApprovalRequestedEvent }
assertNotNull(approval, "ApprovalRequestedEvent must be emitted for PROMPT_USER plane-2 rule on T1 tool")
assertEquals(RiskAction.PROMPT_USER, approval?.riskSummary?.recommendedAction)
job.cancel()
job.join()
}
@Test
fun `null assessor means executor is called normally (regression guard)`(): Unit = runBlocking {
val executor = RecordingExecutor()
val (orchestrator, eventStore) = buildOrchestrator(executor, FakeFileWriteTool(), assessorRule = null)
val sessionId = SessionId("gate-null")
val config = OrchestrationConfig(retryPolicy = RetryPolicy(maxAttempts = 1, backoffMs = 0))
orchestrator.run(sessionId, singleStageGraph(), config)
val events = eventStore.read(sessionId)
assertTrue(executor.executeCalled.get(), "executor must be called when assessor is null")
assertNull(events.find { it.payload is ToolCallAssessedEvent }, "ToolCallAssessedEvent must NOT be emitted when assessor is null")
}
}