feat(validation): auto build-gate on terminal freestyle stage (Gate 4 floor)

Closes the COMPLETED-lie where a freestyle stage passes by producing a
schema-valid artifact regardless of whether its build actually succeeded:
a verify stage could run 'npm run build', watch it fail, then self-attest a
build_verified artifact and transition to done. Tool failures never failed
the stage.

The compiler now flags the terminal stage autoBuildGate=true when no stage
declares an explicit build_expectation. Code-ness is decided at runtime, not
compile time (the declared kind is always the generic file_written; only the
written paths reveal code): runExecutionGate folds the FileWritten manifest via
sessionProducedBuildTarget() and promotes to a PROJECT build when a build
manifest (package_json/gradle_module) or an imports_resolve-bearing path was
written, then runs the real command from .correx/project.toml [commands]. A
non-clean exit fails the stage retryably; a docs-only plan is left untouched.

Live-proven (run 3, session 28812b44): gate fired on the real
'npm --prefix frontend run build', exit 2 on a bad tsconfig, RetryAttempted
instead of a false COMPLETED.

Modules green: transitions+workflow 67, kernel 60.
This commit is contained in:
2026-07-07 01:01:12 +04:00
parent 79e2c38a88
commit 597a26d551
6 changed files with 185 additions and 11 deletions
+7
View File
@@ -13,3 +13,10 @@ test-all = "./gradlew check --rerun-tasks"
test-module = "./gradlew :core:<module>:test --rerun-tasks" test-module = "./gradlew :core:<module>:test --rerun-tasks"
context-lookup = "python scripts/ctx.py <query>" context-lookup = "python scripts/ctx.py <query>"
epic-status = "bash scripts/epic-status.sh" epic-status = "bash scripts/epic-status.sh"
# Build-gate aliases (BuildExpectation → commandAlias): MODULE→typecheck, PROJECT→build, TESTS→test.
# The gate runner is whitespace-split / no-shell and runs at workspace_root (the repo), so `--prefix
# frontend` targets the Node subproject without a `cd`. `npm run build` = `tsc && vite build` (npm's
# own shell handles the &&), which fails on a dangling asset import — exactly the COMPLETED-lie hole.
typecheck = "npm --prefix frontend run build"
build = "npm --prefix frontend run build"
test = "npm --prefix frontend test"
+3
View File
@@ -81,3 +81,6 @@ apps/server/logs/
# local QA scratch workspace (nested git repo) # local QA scratch workspace (nested git repo)
/qa/ /qa/
# QA scaffold artifact (freestyle build-gate runs)
frontend/
@@ -137,6 +137,7 @@ import com.correx.core.tools.registry.ToolRegistry
import com.correx.core.transitions.evaluation.EvaluationContext import com.correx.core.transitions.evaluation.EvaluationContext
import com.correx.core.transitions.evaluation.PromptResolver import com.correx.core.transitions.evaluation.PromptResolver
import com.correx.core.transitions.execution.StageExecutionResult import com.correx.core.transitions.execution.StageExecutionResult
import com.correx.core.transitions.graph.BuildExpectation
import com.correx.core.transitions.graph.StageConfig import com.correx.core.transitions.graph.StageConfig
import com.correx.core.transitions.graph.WorkflowGraph import com.correx.core.transitions.graph.WorkflowGraph
import com.correx.core.transitions.resolution.TransitionDecision import com.correx.core.transitions.resolution.TransitionDecision
@@ -237,6 +238,11 @@ private const val REVIEW_OBJECTIVE_CAP = 4_000
private const val STATIC_ANALYSIS_SUMMARY_CAP = 2_000 private const val STATIC_ANALYSIS_SUMMARY_CAP = 2_000
private const val STATIC_ANALYSIS_FEEDBACK_CAP = 6_000 private const val STATIC_ANALYSIS_FEEDBACK_CAP = 6_000
// KindInference kinds for build manifests — a written file of one of these declares a buildable
// project toolchain even before any source lands, so the auto build gate fires (see
// SessionOrchestrator.sessionProducedBuildTarget). Kept in sync with KindInference's manifest kinds.
private val BUILD_MANIFEST_KINDS = setOf("package_json", "gradle_module")
@SuppressWarnings( @SuppressWarnings(
"ForbiddenComment", "ForbiddenComment",
"UnusedParameter", "UnusedParameter",
@@ -2227,6 +2233,31 @@ abstract class SessionOrchestrator(
* ToolInvocationRequested → this stage's invocation ids, FileWrittenEvent → the paths written * ToolInvocationRequested → this stage's invocation ids, FileWrittenEvent → the paths written
* under them. Pure projection, replay-safe. * under them. Pure projection, replay-safe.
*/ */
/**
* True if the session has committed to a buildable project — it wrote either a real code module
* (a path whose inferred kind carries the `imports_resolve` COMPILER assertion: TS/React/Kotlin
* source) OR a build manifest ([BUILD_MANIFEST_KINDS]: package.json, a Gradle module) that
* declares a toolchain. Read from the recorded FileWritten manifest (invariant #9: replay-safe,
* no filesystem access), it is the run-time signal that warrants the auto build gate — the
* planner's declared kinds (uniformly `file_written`) never reveal this, but the written paths do.
*
* The manifest case is what catches the degenerate scaffold that writes only `package.json` and
* no sources: `npm run build` then fails for lack of an entry, so the gate rejects the empty
* "project" instead of letting it reach COMPLETED. A docs-only plan writes neither, so it is left
* alone.
*/
private fun sessionProducedBuildTarget(sessionId: SessionId): Boolean =
eventStore.read(sessionId)
.mapNotNull { it.payload as? FileWrittenEvent }
.filter { it.postImageHash != null }
.map { it.path }
.distinct()
.any { path ->
val kind = KindInference.kindFor(path) ?: return@any false
kind in BUILD_MANIFEST_KINDS ||
KindContractTable.assertionsFor(kind, path).any { it.id == "imports_resolve" }
}
private fun stageWrittenPaths(sessionId: SessionId, stageId: StageId): List<String> { private fun stageWrittenPaths(sessionId: SessionId, stageId: StageId): List<String> {
val events = eventStore.read(sessionId) val events = eventStore.read(sessionId)
val invocationIds = events.mapNotNull { it.payload as? ToolInvocationRequestedEvent } val invocationIds = events.mapNotNull { it.payload as? ToolInvocationRequestedEvent }
@@ -2313,16 +2344,26 @@ abstract class SessionOrchestrator(
effectives: RunEffectives, effectives: RunEffectives,
profileCommands: Map<String, String>, profileCommands: Map<String, String>,
): StageExecutionResult { ): StageExecutionResult {
val alias = stageConfig.buildExpectation.commandAlias ?: return StageExecutionResult.Success(emptyList()) // An explicit build_expectation resolves directly. Otherwise, if the compiler flagged this
// (terminal) stage as auto-gateable and the session actually wrote a code module, promote to a
// PROJECT build — the LLM planner never sets build_expectation on its file_written stages, so
// this is the only path that gates a scaffold's compilability. Code-ness is judged here (not at
// compile time) from the real FileWritten manifest, so a docs-only plan is left alone.
val expectation: BuildExpectation? = when {
stageConfig.buildExpectation != BuildExpectation.NONE -> stageConfig.buildExpectation
stageConfig.autoBuildGate && sessionProducedBuildTarget(sessionId) -> BuildExpectation.PROJECT
else -> null
}
val alias = expectation?.commandAlias ?: return StageExecutionResult.Success(emptyList())
val runner = staticAnalysisRunner val runner = staticAnalysisRunner
val workspaceRoot = effectives.policy?.workspaceRoot val workspaceRoot = effectives.policy?.workspaceRoot
if (runner == null || workspaceRoot == null) return StageExecutionResult.Success(emptyList()) if (runner == null || workspaceRoot == null) return StageExecutionResult.Success(emptyList())
val command = profileCommands[alias] val command = profileCommands[alias]
if (command.isNullOrBlank()) { if (command.isNullOrBlank()) {
log.warn( log.warn(
"[Orchestrator] stage {} declares build_expectation={} but project profile has no '{}' " + "[Orchestrator] stage {} needs a {} build gate but project profile has no '{}' " +
"command — skipping execution gate", "command — skipping execution gate",
stageId.value, stageConfig.buildExpectation, alias, stageId.value, expectation, alias,
) )
return StageExecutionResult.Success(emptyList()) return StageExecutionResult.Success(emptyList())
} }
@@ -2334,10 +2375,10 @@ abstract class SessionOrchestrator(
log.warn( log.warn(
"[Orchestrator] execution gate failed session={} stage={} expectation={} command={}", "[Orchestrator] execution gate failed session={} stage={} expectation={} command={}",
sessionId.value, stageId.value, stageConfig.buildExpectation, command, sessionId.value, stageId.value, expectation, command,
) )
return StageExecutionResult.Failure( return StageExecutionResult.Failure(
"stage ${stageId.value} did not pass its ${stageConfig.buildExpectation} build gate. " + "stage ${stageId.value} did not pass its $expectation build gate. " +
"Fix these before proceeding:\n\n$ $command (exit ${run.exitCode})\n" + "Fix these before proceeding:\n\n$ $command (exit ${run.exitCode})\n" +
run.output.takeLast(STATIC_ANALYSIS_FEEDBACK_CAP), run.output.takeLast(STATIC_ANALYSIS_FEEDBACK_CAP),
retryable = true, retryable = true,
@@ -43,5 +43,13 @@ data class StageConfig(
// deterministic funnel passes and raises advisory PR-comment findings; a high-confidence // deterministic funnel passes and raises advisory PR-comment findings; a high-confidence
// correctness FAIL can block retryably until the review-block budget is spent. Default off. // correctness FAIL can block retryably until the review-block budget is spent. Default off.
val semanticReview: Boolean = false, val semanticReview: Boolean = false,
// Deterministic build-gate floor (staged-verification §Gate 4). Set by the compiler on the plan's
// terminal stage when no stage declares an explicit [buildExpectation]. The LLM planner emits every
// file-writing stage as the generic `file_written` kind (never a typed code kind) and rarely sets
// build_expectation, so a plan that scaffolds real code would otherwise never hit an execution gate.
// When this is set and NONE is declared, runExecutionGate promotes to a PROJECT build ONLY IF the
// session actually wrote a code module (decided at run time from the FileWritten manifest + path→kind
// inference), so a docs-only plan is left untouched. Explicit build_expectation always wins.
val autoBuildGate: Boolean = false,
val metadata: Map<String, String> = emptyMap(), val metadata: Map<String, String> = emptyMap(),
) )
@@ -44,6 +44,31 @@ class ExecutionPlanCompiler(
if (plan.stages.isEmpty()) throw WorkflowValidationException("execution_plan has no stages") if (plan.stages.isEmpty()) throw WorkflowValidationException("execution_plan has no stages")
validateTools(plan) validateTools(plan)
// Parse + validate every stage's declared build_expectation up front — also feeds the
// deterministic build-gate guarantee below.
val declaredExpectations = plan.stages.associate { s ->
s.id to (
BuildExpectation.fromPlan(s.buildExpectation)
?: throw WorkflowValidationException(
"stage '${s.id}' has invalid build_expectation '${s.buildExpectation}' — " +
"use one of none|module|project|tests",
)
)
}
// Deterministic build-gate floor. Code kinds carry an `imports_resolve` COMPILER-layer
// contract assertion, but COMPILER assertions are enforced only by the execution gate, which
// fires only when a stage sets build_expectation. The LLM planner emits every file-writing
// stage as the generic `file_written` kind (never a typed code kind) and rarely sets
// build_expectation, so a scaffold with a dangling import (`import './index.css'` for a file
// never written) sails through to COMPLETED. Close it: when no stage declares any gate, flag
// the terminal stage for an auto build gate — placed there, not per-stage, so it runs when the
// project is whole rather than failing legitimately-incomplete intermediate stages. Whether it
// ACTUALLY builds is decided at run time (SessionOrchestrator.runExecutionGate) from the real
// FileWritten manifest — the plan's declared kinds don't reveal code-ness, but the written
// paths do — so a docs-only plan is left alone and a code plan is always gated.
val autoGateStageId: String? =
if (declaredExpectations.values.all { it == BuildExpectation.NONE }) terminalStageId(plan) else null
val stageMap = plan.stages.associate { s -> val stageMap = plan.stages.associate { s ->
// `produces` is the unique slot name referenced by needs/edges; `kind` selects the // `produces` is the unique slot name referenced by needs/edges; `kind` selects the
// registered artifact kind. Plans predating the split used one string for both. // registered artifact kind. Plans predating the split used one string for both.
@@ -59,11 +84,6 @@ class ExecutionPlanCompiler(
staticManifest = emptyList(), staticManifest = emptyList(),
planDerived = PlanDerivedManifest.deriveAllowedWrites(s), planDerived = PlanDerivedManifest.deriveAllowedWrites(s),
).sorted() ).sorted()
val buildExpectation = BuildExpectation.fromPlan(s.buildExpectation)
?: throw WorkflowValidationException(
"stage '${s.id}' has invalid build_expectation '${s.buildExpectation}' — " +
"use one of none|module|project|tests",
)
StageId(s.id) to StageConfig( StageId(s.id) to StageConfig(
produces = listOf(TypedArtifactSlot(name = ArtifactId(s.produces), kind = kind)), produces = listOf(TypedArtifactSlot(name = ArtifactId(s.produces), kind = kind)),
needs = s.needs.map { ArtifactId(it) }.toSet(), needs = s.needs.map { ArtifactId(it) }.toSet(),
@@ -73,7 +93,8 @@ class ExecutionPlanCompiler(
// the contract gate will require to exist. Globs (src/**) are write-permission scopes, // the contract gate will require to exist. Globs (src/**) are write-permission scopes,
// not existence guarantees, so they are excluded here. // not existence guarantees, so they are excluded here.
expectedFiles = s.writes.map { it.trim() }.filter { it.isNotEmpty() && !isGlob(it) }, expectedFiles = s.writes.map { it.trim() }.filter { it.isNotEmpty() && !isGlob(it) },
buildExpectation = buildExpectation, buildExpectation = declaredExpectations.getValue(s.id),
autoBuildGate = s.id == autoGateStageId,
semanticReview = s.semanticReview, semanticReview = s.semanticReview,
tokenBudget = DEFAULT_STAGE_TOKEN_BUDGET, tokenBudget = DEFAULT_STAGE_TOKEN_BUDGET,
metadata = mapOf("promptInline" to s.prompt), metadata = mapOf("promptInline" to s.prompt),
@@ -146,6 +167,11 @@ class ExecutionPlanCompiler(
} }
} }
// The stage the plan ends on — the one whose edge points at "done" (a well-formed plan has exactly
// one). Falls back to the last declared stage if no terminal edge is present yet.
private fun terminalStageId(plan: ExecutionPlanModel): String =
plan.edges.firstOrNull { it.to == TERMINAL }?.from ?: plan.stages.last().id
private fun isGlob(path: String): Boolean = path.any { it == '*' || it == '?' || it == '[' } private fun isGlob(path: String): Boolean = path.any { it == '*' || it == '?' || it == '[' }
private fun validateReachability(declared: Set<String>, transitions: Set<TransitionEdge>, start: StageId) { private fun validateReachability(declared: Set<String>, transitions: Set<TransitionEdge>, start: StageId) {
@@ -4,6 +4,7 @@ import com.correx.core.artifacts.kind.ConfigArtifactKind
import com.correx.core.artifacts.kind.DefaultArtifactKindRegistry import com.correx.core.artifacts.kind.DefaultArtifactKindRegistry
import com.correx.core.artifacts.kind.JsonSchema import com.correx.core.artifacts.kind.JsonSchema
import com.correx.core.artifacts.kind.JsonSchemaProperty import com.correx.core.artifacts.kind.JsonSchemaProperty
import com.correx.core.events.types.StageId
import org.junit.jupiter.api.Test import org.junit.jupiter.api.Test
import org.junit.jupiter.api.assertThrows import org.junit.jupiter.api.assertThrows
import kotlin.test.assertEquals import kotlin.test.assertEquals
@@ -362,4 +363,92 @@ class ExecutionPlanCompilerTest {
) )
assertThrows<WorkflowValidationException> { compiler.compile(bad, "wf") } assertThrows<WorkflowValidationException> { compiler.compile(bad, "wf") }
} }
// --- deterministic build-gate guarantee -------------------------------------------------------
// Registry whose kinds resolve to code contracts (react_component / react_entry carry the
// `imports_resolve` COMPILER assertion in KindContractTable).
private val codeRegistry = DefaultArtifactKindRegistry().also {
it.register(ConfigArtifactKind(id = "react_component", schema = trivialSchema, llmEmitted = false))
it.register(ConfigArtifactKind(id = "react_entry", schema = trivialSchema, llmEmitted = false))
}
private val codeCompiler = ExecutionPlanCompiler(codeRegistry)
private val codePlanNoGate = """
{
"goal": "scaffold a react app",
"stages": [
{ "id": "entry", "prompt": "write main.tsx", "produces": "app_entry", "kind": "react_entry",
"needs": [], "tools": ["file_write"] },
{ "id": "views", "prompt": "write components", "produces": "app_views", "kind": "react_component",
"needs": ["app_entry"], "tools": ["file_write"] }
],
"edges": [
{ "from": "entry", "to": "views", "condition": { "type": "always_true" } },
{ "from": "views", "to": "done", "condition": { "type": "always_true" } }
]
}
""".trimIndent()
@Test
fun `terminal stage of an ungated plan is flagged for an auto build gate`() {
val graph = codeCompiler.compile(codePlanNoGate, "wf")
val views = graph.stages[StageId("views")]!!
assertTrue(
views.autoBuildGate,
"terminal stage of an ungated plan must be flagged for an auto build gate",
)
assertEquals(
com.correx.core.transitions.graph.BuildExpectation.NONE,
views.buildExpectation,
"the flag alone — declared expectation stays NONE; the runtime promotes to PROJECT only if a " +
"code module was actually written",
)
assertTrue(
!graph.stages[StageId("entry")]!!.autoBuildGate,
"intermediate stages are not flagged — the gate runs when the project is whole, not per-stage",
)
}
@Test
fun `an explicitly declared gate suppresses the auto gate`() {
// Planner set a MODULE gate on the entry stage; the compiler must not add its own.
val planned = """
{
"goal": "scaffold a react app",
"stages": [
{ "id": "entry", "prompt": "write main.tsx", "produces": "app_entry", "kind": "react_entry",
"needs": [], "tools": ["file_write"], "build_expectation": "module" },
{ "id": "views", "prompt": "write components", "produces": "app_views", "kind": "react_component",
"needs": ["app_entry"], "tools": ["file_write"] }
],
"edges": [
{ "from": "entry", "to": "views", "condition": { "type": "always_true" } },
{ "from": "views", "to": "done", "condition": { "type": "always_true" } }
]
}
""".trimIndent()
val graph = codeCompiler.compile(planned, "wf")
assertEquals(
com.correx.core.transitions.graph.BuildExpectation.MODULE,
graph.stages[StageId("entry")]!!.buildExpectation,
)
assertTrue(
graph.stages.values.none { it.autoBuildGate },
"no stage is auto-flagged when the planner already declares a gate anywhere",
)
}
@Test
fun `exactly the terminal stage of an ungated plan is auto-flagged`() {
// The compiler flags the terminal stage regardless of the produced kinds — code-ness is a
// run-time decision (SessionOrchestrator reads the FileWritten manifest), so a non-code plan
// is flagged too but the runtime build gate then skips it.
val graph = compiler.compile(validPlan, "wf")
assertEquals(
1,
graph.stages.values.count { it.autoBuildGate },
"exactly the single terminal stage of an ungated plan is flagged for the auto build gate",
)
}
} }