fix: require real progress evidence artifacts

hooks/force-recall/handler.ts

@@ -96,13 +96,36 @@ async function runLongTaskWrapper(workspaceDir: string, ctx: any): Promise<any |
 }
 
 function buildProgressEvidence(wrapperResult: any): Record<string, unknown> | null {
+  const candidate = wrapperResult?.progressEvidence;
+  if (!candidate || typeof candidate !== "object" || Array.isArray(candidate)) {
+    return null;
+  }
+
   const progressEvidence: Record<string, unknown> = {};
 
-  const taskName = typeof wrapperResult?.taskRecord?.task_name === "string"
+  const sessionKey = typeof candidate.sessionKey === "string"
-    ? wrapperResult.taskRecord.task_name.trim()
+    ? candidate.sessionKey.trim()
     : "";
-  if (wrapperResult?.silentLaunchOk === true && taskName) {
+  if (sessionKey) {
-    progressEvidence.sessionKey = taskName;
+    progressEvidence.sessionKey = sessionKey;
+  }
+
+  const runId = typeof candidate.runId === "string"
+    ? candidate.runId.trim()
+    : "";
+  if (runId) {
+    progressEvidence.runId = runId;
+  }
+
+  if (Array.isArray(candidate.modified_files) && candidate.modified_files.length > 0) {
+    progressEvidence.modified_files = candidate.modified_files;
+  }
+
+  const verificationResult = typeof candidate.verificationResult === "string"
+    ? candidate.verificationResult.trim()
+    : "";
+  if (verificationResult) {
+    progressEvidence.verificationResult = verificationResult;
   }
 
   return Object.keys(progressEvidence).length > 0 ? progressEvidence : null;
@@ -174,9 +197,8 @@ function buildGateLockInput(wrapperResult: any): Record<string, unknown> {
   const progressEvidenceReason = claimedProgression && !progressEvidence
     ? "progression claim requires concrete evidence such as sessionKey, runId, modified_files, or verification result"
     : "";
-  const hasExternalizedCheckpointEvidence = wrapperResult.silentLaunchOk === true
+  const hasExternalizedCheckpointEvidence = typeof wrapperResult.externalizedCheckpointPath === "string"
-    && typeof wrapperResult.taskRecord?.task_name === "string"
+    && wrapperResult.externalizedCheckpointPath.trim().length > 0;
-    && wrapperResult.taskRecord.task_name.trim().length > 0;
   const hasButtonPathClosureEvidence = needsOwnerDecision && wrapperResult.silentLaunchOk === true;
 
   return {
@@ -206,8 +228,8 @@ function buildGateLockInput(wrapperResult: any): Record<string, unknown> {
     dispatchEvidence: "",
     fileChangeEvidence: "",
     verificationEvidence: "",
-    checkpointArtifactEvidence: hasExternalizedCheckpointEvidence ? wrapperResult.taskRecord.task_name.trim() : "",
+    checkpointArtifactEvidence: hasExternalizedCheckpointEvidence ? wrapperResult.externalizedCheckpointPath.trim() : "",
-    externalizedCheckpointPath: hasExternalizedCheckpointEvidence ? wrapperResult.taskRecord.task_name.trim() : "",
+    externalizedCheckpointPath: hasExternalizedCheckpointEvidence ? wrapperResult.externalizedCheckpointPath.trim() : "",
     externalizedTrigger: hasExternalizedCheckpointEvidence ? "hook-preflight-checkpoint" : "",
     handoffMode: hasButtonPathClosureEvidence ? (wrapperResult.handoff?.mode ?? "button_path") : "direct_reply",
     replyClosureMode: hasButtonPathClosureEvidence ? (wrapperResult.handoff?.mode ?? "button_path") : "direct_reply",

scripts/test_force_recall_long_task_preflight.mjs

@@ -209,7 +209,8 @@ async function main() {
       dispatched: true,
       event: 'dispatch',
     },
-    taskRecord: { task_name: 'task-123' },
+    progressEvidence: { sessionKey: 'task-123' },
+    externalizedCheckpointPath: 'checkpoints/task-123.json',
     handoff: { mode: 'direct_reply' },
   }), async () => runScenario(forceRecall, requestText));
   assert.match(passInjected, /gateStatus=pass/, 'hook pass-path should pass when wrapper provides concrete progressEvidence');
@@ -285,7 +286,8 @@ async function main() {
       dispatched: true,
       event: 'dispatch',
     },
-    taskRecord: { task_name: 'task-spec-review-missing-evidence' },
+    progressEvidence: { sessionKey: 'task-spec-review-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-spec-review-missing-evidence.json',
     handoff: { mode: 'direct_reply' },
   }), async () => runScenario(forceRecall, plannerOnlyRequestText));
   assert.match(specReviewWithoutEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook spec-review missing-evidence path should emit auto-chain plan block');
@@ -310,7 +312,8 @@ async function main() {
       dispatched: true,
       event: 'dispatch',
     },
-    taskRecord: { task_name: 'task-fix-slice-missing-evidence' },
+    progressEvidence: { sessionKey: 'task-fix-slice-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-fix-slice-missing-evidence.json',
     handoff: { mode: 'direct_reply' },
   }), async () => runScenario(forceRecall, plannerOnlyRequestText));
   assert.match(fixSliceWithoutEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook fix-slice missing-evidence path should emit auto-chain plan block');
@@ -334,7 +337,8 @@ async function main() {
       dispatched: true,
       event: 'dispatch',
     },
-    taskRecord: { task_name: 'task-implementation-missing-evidence' },
+    progressEvidence: { sessionKey: 'task-implementation-missing-evidence' },
+    externalizedCheckpointPath: 'checkpoints/task-implementation-missing-evidence.json',
     handoff: { mode: 'direct_reply' },
   }), async () => runScenario(forceRecall, plannerOnlyRequestText));
   assert.match(specReviewWithoutImplementationEvidenceInjected, /\[LONG_TASK_AUTO_CHAIN_PLAN\]/, 'hook implementation missing-evidence path should emit auto-chain plan block');