omnius 1.0.878 → 1.0.879

package/dist/library.js

@@ -8250,6 +8250,22 @@ var init_network_egress_policy = __esm({
   }
 });
 
+// packages/execution/dist/shell-runs.js
+var init_shell_runs = __esm({
+  "packages/execution/dist/shell-runs.js"() {
+    "use strict";
+  }
+});
+
+// packages/execution/dist/shell-run-control.js
+var init_shell_run_control = __esm({
+  "packages/execution/dist/shell-run-control.js"() {
+    "use strict";
+    init_process_lifecycle();
+    init_process_kill();
+  }
+});
+
 // packages/execution/dist/tools/log-packet.js
 var init_log_packet = __esm({
   "packages/execution/dist/tools/log-packet.js"() {
@@ -8569,12 +8585,14 @@ function exactTrailingExitReporterPrimary(command) {
 }
 function trustedTrailingExitReporterPrimary(command, receipt) {
   const primary = exactTrailingExitReporterPrimary(command);
-  if (!primary || !receipt || receipt.timedOut || receipt.exitCode !== 0 || receipt.primaryExitCode !== 0 || receipt.primaryExitProvenance !== "trailing_exit_reporter" || normalizeShellCommand(receipt.primaryCommand ?? "") !== normalizeShellCommand(primary)) {
+  if (!primary || !receipt || receipt.timedOut || receipt.cancelled === true || receipt.exitCode !== 0 || receipt.primaryExitCode !== 0 || receipt.primaryExitProvenance !== "trailing_exit_reporter" || normalizeShellCommand(receipt.primaryCommand ?? "") !== normalizeShellCommand(primary)) {
     return null;
   }
   return normalizeShellCommand(receipt.canonicalCommand) === normalizeShellCommand(command) ? primary : null;
 }
 function describeUntrustedVerificationCommand(command, receipt) {
+  if (receipt?.cancelled === true)
+    return "the command belongs to a cancelled execution";
   const primary = exactTrailingExitReporterPrimary(command);
   if (primary) {
     if (!receipt) {
@@ -8736,7 +8754,7 @@ function isTeeSink(stage2) {
   return /^tee(?:\s+(?:-a|--append|--ignore-interrupts))*\s+(?:--\s+)?\S+/i.test(stage2.trim());
 }
 function receiptTrustsPipeline(receipt) {
-  return Boolean(receipt && receipt.pipefail === true && receipt.timedOut !== true && receipt.exitCode === 0);
+  return Boolean(receipt && receipt.pipefail === true && receipt.timedOut !== true && receipt.cancelled !== true && receipt.exitCode === 0);
 }
 function isFileDescriptorRedirectionAmpersand(command, index) {
   return command[index - 1] === ">" && /[0-9-]/.test(command[index + 1] ?? "");
@@ -8898,6 +8916,8 @@ function commandIsPureReadOnlyPipeline(command) {
   });
 }
 function commandReliablySatisfiesVerifyCommand(observedCommand, verifyCommand, receipt) {
+  if (receipt?.cancelled === true)
+    return false;
   if (/[\r\n]/.test(observedCommand))
     return false;
   const reporterPrimary = trustedTrailingExitReporterPrimary(observedCommand, receipt);
@@ -9002,6 +9022,8 @@ function parseVerificationStage(source) {
   return { source, words: words3, writesRedirectedFile };
 }
 function parseVerificationCommandStages(command, receipt) {
+  if (receipt?.cancelled === true)
+    return null;
   const source = trustedTrailingExitReporterPrimary(command, receipt) ?? command;
   if (!source.trim() || hasUnsafeVerifierSyntaxExceptPipeline(source))
     return null;
@@ -9337,6 +9359,8 @@ var init_system_auth = __esm({
 var init_shell = __esm({
   "packages/execution/dist/tools/shell.js"() {
     "use strict";
+    init_shell_runs();
+    init_shell_run_control();
     init_process_kill();
     init_process_lifecycle();
     init_log_packet();
@@ -82736,7 +82760,8 @@ var init_git_info = __esm({
 var init_background_task = __esm({
   "packages/execution/dist/tools/background-task.js"() {
     "use strict";
-    init_process_kill();
+    init_shell_runs();
+    init_shell_run_control();
     init_background_process_receipt();
     init_process_lifecycle();
   }
@@ -88393,6 +88418,7 @@ function evidenceFingerprint(evidence) {
       cwd: evidence.commandReceipt.cwd,
       exitCode: evidence.commandReceipt.exitCode,
       timedOut: evidence.commandReceipt.timedOut,
+      cancelled: evidence.commandReceipt.cancelled,
       startedAtIso: evidence.commandReceipt.startedAtIso,
       finishedAtIso: evidence.commandReceipt.finishedAtIso
     } : void 0,
@@ -88407,7 +88433,7 @@ function normalizeCommandText(value) {
   return value.replace(/\s+/g, " ").trim();
 }
 function isSuccessfulCommandEvidenceReceipt(receipt) {
-  return receipt.schema === "omnius.command-evidence-receipt.v1" && receipt.timedOut === false && receipt.exitCode === 0 && (receipt.primaryExitCode === void 0 || receipt.primaryExitCode === 0) && (receipt.runnerExitCode === void 0 || receipt.runnerExitCode === 0);
+  return receipt.schema === "omnius.command-evidence-receipt.v1" && receipt.timedOut === false && receipt.cancelled !== true && receipt.exitCode === 0 && (receipt.primaryExitCode === void 0 || receipt.primaryExitCode === 0) && (receipt.runnerExitCode === void 0 || receipt.runnerExitCode === 0);
 }
 function normalizeTrustedCommandReceipt(receipt) {
   if (receipt?.schema !== "omnius.command-evidence-receipt.v1") {
@@ -88427,7 +88453,7 @@ function normalizeTrustedCommandReceipt(receipt) {
   if (finishedAt < startedAt) {
     throw new WorkboardError("invalid_command_receipt", "command receipt finished before it started");
   }
-  if (typeof receipt.pipefail !== "boolean" || typeof receipt.timedOut !== "boolean") {
+  if (typeof receipt.pipefail !== "boolean" || typeof receipt.timedOut !== "boolean" || receipt.cancelled !== void 0 && typeof receipt.cancelled !== "boolean") {
     throw new WorkboardError("invalid_command_receipt", "command evidence receipt is missing process flags");
   }
   return cloneCommandEvidenceReceipt({
@@ -89059,6 +89085,7 @@ function writeTodosLocked(sessionId, incoming) {
       completedAt: status2 === "completed" ? existing?.completedAt ?? now2 : void 0,
       verifyCommand: t.verifyCommand ?? existing?.verifyCommand,
       // REG-37
+      completionJustification: t.completionJustification ?? existing?.completionJustification,
       declaredArtifacts: t.declaredArtifacts ?? existing?.declaredArtifacts
       // REG-38
     };
@@ -336547,6 +336574,7 @@ var init_dist5 = __esm({
     init_provenance();
     init_network_egress_policy();
     init_shell();
+    init_shell_runs();
     init_debate();
     init_replay_with_intervention();
     init_file_read();
@@ -343499,13 +343527,13 @@ function createOpenCodeCompatibilityTools() {
     type: "function",
     function: {
       name: "shell",
-      description: "Execute a shell command and return its output. Commands run on Linux using /bin/bash. Quote file paths containing spaces or special characters. Prefer dedicated tools over shell commands when possible. When output is large, the full result is saved to a file and a truncated preview is returned. Rely on automatic truncation unless filtering the output is more useful. Commands accept an optional timeout, background commands have no timeout by default. Background commands return immediately, and you will be notified when they complete.",
+      description: "Execute a shell command and return its output. Commands run on Linux using /bin/bash. Quote file paths containing spaces or special characters. Prefer dedicated tools over shell commands when possible. When output is large, the full result is saved to a file and a truncated preview is returned. Every call must explicitly choose a positive integer timeout in milliseconds or null for no deadline. Background commands return immediately, and you will be notified when they complete.",
       parameters: { type: "object", properties: {
         command: { type: "string", description: "Shell command string to execute" },
         workdir: { type: "string", description: "Working directory to execute the command in. Defaults to the current working directory. When possible, avoid changing directories in the command and set the working directory here instead." },
-        timeout: { type: "integer", minimum: 0, description: "Timeout in milliseconds. Set to 0 to disable the timeout. Defaults to 120000 for foreground commands. Background commands have no timeout by default." },
+        timeout: { type: ["integer", "null"], minimum: 1, description: "Required positive integer deadline in milliseconds, or null for no deadline until an explicit Stop." },
         background: { type: "boolean", description: "Run the command in the background and return immediately (useful for dev servers and long-running builds). You do not need to use '&' at the end of the command when using this parameter. You will be notified when it completes. DO NOT poll for completion." }
-      }, required: ["command"], additionalProperties: false }
+      }, required: ["command", "timeout"], additionalProperties: false }
     }
   }];
 }
@@ -350494,10 +350522,10 @@ function hasClaimPathOverlap(claim, evidence) {
 }
 function trustedSuccessfulReceipt(evidence) {
   const receipt = evidence.receipt;
-  return evidence.success === true && Boolean(receipt && receipt.canonicalCommand && receipt.cwd && receipt.exitCode === 0 && receipt.timedOut !== true && receipt.finishedAtIso);
+  return evidence.success === true && Boolean(receipt && receipt.canonicalCommand && receipt.cwd && receipt.exitCode === 0 && receipt.timedOut !== true && receipt.cancelled !== true && receipt.finishedAtIso);
 }
 function evidenceMatchesClaimNeed(claim, evidence, need) {
-  if (evidence.success !== true)
+  if (evidence.success !== true || evidence.receipt?.cancelled === true)
     return false;
   const pathOverlap = hasClaimPathOverlap(claim, evidence);
   switch (need.evidenceKind) {
@@ -350555,7 +350583,7 @@ function reconcileClaimsWithEvidence(ledger) {
     const allNeedsMet = claim.evidenceNeeds.every((need) => linked.some((entry) => evidenceMatchesClaimNeed(claim, entry, need)) || // Preserve explicit evidence links authored by a caller. They remain
     // subject to failure/receipt validation above, but should not be
     // reclassified only because a legacy emitter lacked role metadata.
-    explicit.some((entry) => entry.success === true));
+    explicit.some((entry) => entry.success === true && entry.receipt?.cancelled !== true));
     const independentMet = !claim.requiresIndependentEvidence || Boolean(independent);
     return {
       ...claim,
@@ -350661,7 +350689,7 @@ function isSuccessfulVerificationEvidence(entry) {
     return false;
   if (!entry.receipt)
     return true;
-  return entry.receipt.schema === "omnius.command-evidence-receipt.v1" && Boolean(entry.receipt.canonicalCommand.trim()) && Boolean(entry.receipt.cwd.trim()) && entry.receipt.exitCode === 0 && !entry.receipt.timedOut && entry.receipt.declaredVerification?.outcome !== "fail";
+  return entry.receipt.schema === "omnius.command-evidence-receipt.v1" && Boolean(entry.receipt.canonicalCommand.trim()) && Boolean(entry.receipt.cwd.trim()) && entry.receipt.exitCode === 0 && !entry.receipt.timedOut && entry.receipt.cancelled !== true && entry.receipt.declaredVerification?.outcome !== "fail";
 }
 function isFailedVerificationEvidence(entry) {
   return entry.role === "verification" && (entry.success === false || // A typed failed, timed-out or malformed receipt overrides a positive
@@ -350671,6 +350699,8 @@ function isFailedVerificationEvidence(entry) {
 function isSuccessfulPostMutationObservation(entry) {
   if (entry.success !== true)
     return false;
+  if (entry.receipt?.cancelled === true)
+    return false;
   if (entry.role === "blocker")
     return false;
   if (isMutationEvidence(entry))
@@ -351424,6 +351454,8 @@ function commandIdentity(entry) {
 }
 function checkOutcome(entry) {
   const receipt = entry.receipt;
+  if (receipt?.cancelled === true)
+    return "failed";
   if (receipt?.timedOut === true)
     return "timed_out";
   const start = timestamp(receipt?.startedAtIso);
@@ -355728,6 +355760,11 @@ var CommandEvidenceReceiptSchema = z18.object({
   timedOut: z18.boolean(),
   startedAtIso: timestamp2,
   finishedAtIso: timestamp2,
+  shellRunId: z18.string().uuid().optional(),
+  timeoutMs: integer3.positive().nullable().optional(),
+  cancelled: z18.boolean().optional(),
+  signal: token2.nullable().optional(),
+  observation: z18.object({ kind: z18.literal("search_no_match") }).strict().optional(),
   outputRef: token2.optional(),
   observedPaths: z18.array(token2).optional(),
   artifactHashes: z18.array(z18.object({ path: token2, sha256: z18.string().regex(/^[a-f0-9]{64}$/) }).strict()).optional(),
@@ -357135,6 +357172,7 @@ var commandReceiptSchema = z20.object({
   cwd: token4,
   pipefail: z20.boolean(),
   timedOut: z20.boolean(),
+  cancelled: z20.boolean().optional(),
   startedAtIso: z20.string().datetime(),
   finishedAtIso: z20.string().datetime(),
   exitCode: z20.number().int().nullable().optional(),
@@ -358157,7 +358195,7 @@ var FeatureWorkflowController = class {
       if (binding.kind === "verification") {
         if (result.commandReceipt)
           commandReceipt = commandReceiptSchema.parse(result.commandReceipt);
-        successful = successful && !!commandReceipt && commandReceipt.exitCode === 0 && commandReceipt.timedOut !== true && resolve26(commandReceipt.cwd) === resolve26(this.options.workingDir, card.verifier.cwd ?? ".") && commandVerifiesFinalState(commandReceipt.canonicalCommand, card.verifier.command, commandReceipt) && workspaceRevision === operation.workspaceRevision && state.mutationSequence === operation.mutationSequence;
+        successful = successful && !!commandReceipt && commandReceipt.exitCode === 0 && commandReceipt.timedOut !== true && commandReceipt.cancelled !== true && resolve26(commandReceipt.cwd) === resolve26(this.options.workingDir, card.verifier.cwd ?? ".") && commandVerifiesFinalState(commandReceipt.canonicalCommand, card.verifier.command, commandReceipt) && workspaceRevision === operation.workspaceRevision && state.mutationSequence === operation.mutationSequence;
         if (!successful)
           blockers.push(`Run the declared verifier against the current integrated source: ${card.verifier.command}`);
       }
@@ -360369,7 +360407,7 @@ function evidenceLabel(evidence) {
 }
 function evidenceIsFreshVerification(evidence) {
   const receipt = evidence.receipt;
-  return evidence.success === true && evidence.role === "verification" && Boolean(receipt && receipt.canonicalCommand && receipt.cwd && receipt.exitCode === 0 && receipt.timedOut !== true && receipt.finishedAtIso);
+  return evidence.success === true && evidence.role === "verification" && Boolean(receipt && receipt.canonicalCommand && receipt.cwd && receipt.exitCode === 0 && receipt.timedOut !== true && receipt.cancelled !== true && receipt.finishedAtIso);
 }
 function claimRequiresClosure(claim) {
   return claim.materiality === "high" || claim.requiresIndependentEvidence || claim.attainedState === "hardware_integrated" || claim.attainedState === "hardware_verified";
@@ -360484,7 +360522,7 @@ function buildIntegrationClosure(input) {
     });
   }
   const activeFailures = [...input.activeFailures ?? []];
-  const evidence = (ledger?.evidence ?? []).slice(-MAX_EVIDENCE);
+  const evidence = (ledger?.evidence ?? []).slice(-MAX_EVIDENCE).map((item) => item.receipt?.cancelled === true ? { ...item, success: false } : item);
   for (const item of evidence) {
     const evidenceId = nodeId("evidence", item.id);
     addNode({
@@ -362006,6 +362044,8 @@ function claimAuthority(claim) {
   return claim.source === "derived" ? "inferred" : "adopted";
 }
 function evidenceRole(evidence) {
+  if (evidence.receipt?.cancelled === true)
+    return "blocker";
   if (evidence.role === "verification")
     return "verification";
   if (evidence.role === "blocker" || evidence.success === false)
@@ -362039,7 +362079,7 @@ function freshTodoVerificationEvidence(todo, ledger) {
   const lastMutation = relevantMutationTimes.length > 0 ? Math.max(...relevantMutationTimes) : Number.NEGATIVE_INFINITY;
   const expected = canonicalCommand(todo.verifyCommand);
   return ledger.evidence.filter((entry) => {
-    if (entry.role !== "verification" || entry.success !== true || !entry.receipt || entry.receipt.exitCode !== 0 || entry.receipt.timedOut) {
+    if (entry.role !== "verification" || entry.success !== true || !entry.receipt || entry.receipt.exitCode !== 0 || entry.receipt.timedOut || entry.receipt.cancelled === true) {
       return false;
     }
     if (canonicalCommand(entry.receipt.canonicalCommand) !== expected)
@@ -362070,7 +362110,7 @@ function freshTodoCompletionEvidence(todo, ledger) {
       return false;
     if (entry.role !== "verification" && entry.role !== "delivery")
       return false;
-    if (entry.receipt && (entry.receipt.exitCode !== 0 || entry.receipt.timedOut)) {
+    if (entry.receipt && (entry.receipt.exitCode !== 0 || entry.receipt.timedOut || entry.receipt.cancelled === true)) {
       return false;
     }
     if (!entry.receipt && entry.kind !== "delivery_result" && entry.kind !== "human_input") {
@@ -362881,7 +362921,7 @@ function buildWorkGraphProjectionPatch(input) {
         nodeId: scopedId("evidence-node", evidence.id),
         kind: "evidence",
         label: evidence.summary.slice(0, 1e3) || evidence.id,
-        status: evidence.success === false ? "invalid" : "complete",
+        status: evidence.success === false || evidence.receipt?.cancelled === true ? "invalid" : "complete",
         authority: "adopted",
         bindingIds: [bindingId],
         evidenceRefIds: [evidenceId],
@@ -363733,6 +363773,8 @@ function reconcileCompletedTodosWithEvidence(input) {
       todo.blocker = blocker2;
       continue;
     }
+    if (todo.completionJustification?.trim())
+      continue;
     const startTurn = todo.id && input.todoStartTurnById?.has(todo.id) ? input.todoStartTurnById.get(todo.id) : 0;
     const relevantCalls = input.toolCallLog.filter((call2) => (call2.turn ?? 0) >= startTurn);
     const failedIndex = relevantCalls.findIndex((call2, index2) => call2.success === false && todoMentionsFailedCall(todo, call2) && !hasLaterSuccessForCompatibleEvidence(relevantCalls, index2));
@@ -364201,7 +364243,7 @@ init_dist5();
 // packages/orchestrator/dist/taskValidationEvidence.js
 function successfulTaskValidationExecution(input) {
   const command = input.command.trim();
-  return input.success && input.runtimeAuthored !== true && Boolean(command) && (!input.executionReceipt || input.executionReceipt.timedOut !== true && input.executionReceipt.exitCode === 0 && (input.executionReceipt.primaryExitCode === void 0 || input.executionReceipt.primaryExitCode === 0) && normalizeShellCommand(input.executionReceipt.canonicalCommand) === normalizeShellCommand(command));
+  return input.success && input.runtimeAuthored !== true && Boolean(command) && (!input.executionReceipt || input.executionReceipt.cancelled !== true && input.executionReceipt.timedOut !== true && input.executionReceipt.exitCode === 0 && (input.executionReceipt.primaryExitCode === void 0 || input.executionReceipt.primaryExitCode === 0) && normalizeShellCommand(input.executionReceipt.canonicalCommand) === normalizeShellCommand(command));
 }
 function selectTaskValidationEvidence(input) {
   const command = input.command.trim();
@@ -364319,7 +364361,7 @@ function captureCompletionRuntimeProvenance(capturedAtIso = (/* @__PURE__ */ new
 }
 function isValidObservationReceipt(entry) {
   const receipt = entry.receipt;
-  return entry.success === true && entry.role === "observation" && receipt?.schema === "omnius.command-evidence-receipt.v1" && Boolean(receipt.canonicalCommand.trim()) && Boolean(receipt.cwd.trim()) && receipt.exitCode === 0 && receipt.timedOut === false;
+  return entry.success === true && entry.role === "observation" && receipt?.schema === "omnius.command-evidence-receipt.v1" && Boolean(receipt.canonicalCommand.trim()) && Boolean(receipt.cwd.trim()) && receipt.exitCode === 0 && receipt.timedOut === false && receipt.cancelled !== true;
 }
 function selectedEvidence(ledger) {
   if (!ledger) {
@@ -364330,7 +364372,7 @@ function selectedEvidence(ledger) {
       artifactHashes: []
     };
   }
-  const candidates = ledger.evidence.filter((entry) => entry.success === true && (entry.role === "verification" || entry.role === "mutation" || entry.role === "delivery" && Boolean(entry.deliveryReceipt) || isValidObservationReceipt(entry))).slice(-16);
+  const candidates = ledger.evidence.filter((entry) => entry.success === true && entry.receipt?.cancelled !== true && (entry.role === "verification" || entry.role === "mutation" || entry.role === "delivery" && Boolean(entry.deliveryReceipt) || isValidObservationReceipt(entry))).slice(-16);
   const artifacts = /* @__PURE__ */ new Map();
   const deliveries = /* @__PURE__ */ new Map();
   const commandReceipts = [];
@@ -364729,7 +364771,7 @@ function latestVerification(requirement, mappedEvidenceIds, evidence) {
   const selected = candidates.at(-1)?.entry;
   if (!selected?.receipt)
     return { status: "unverified" };
-  const passed = selected.success === true && selected.receipt.exitCode === 0 && selected.receipt.timedOut === false && selected.receipt.declaredVerification?.outcome !== "fail";
+  const passed = selected.success === true && selected.receipt.exitCode === 0 && selected.receipt.timedOut === false && selected.receipt.cancelled !== true && selected.receipt.declaredVerification?.outcome !== "fail";
   return {
     status: passed ? "satisfied" : "failed",
     projection: {
@@ -365557,7 +365599,9 @@ function reconcileShellExecutionOutcome(command, result) {
   if (!result.success || !command.trim() || !receipt || receipt.schema !== "omnius.command-evidence-receipt.v1" || typeof receipt.canonicalCommand !== "string" || normalizeShellCommand(receipt.canonicalCommand) !== normalizeShellCommand(command))
     return result;
   let error;
-  if (receipt.timedOut)
+  if (receipt.cancelled)
+    error = "Command execution stopped (recorded execution receipt).";
+  else if (receipt.timedOut)
     error = "Command execution timed out (recorded execution receipt).";
   else {
     const candidate = (value) => value.kind === "validation" || value.kind === "other" && value.substantiveStageCount === 1;
@@ -366606,7 +366650,7 @@ function evaluateDeliveryCoverage(manifest, availableEvidenceIds = [], sourceObs
         if (!acceptance.requiredFor.includes(claim.state))
           continue;
         const expected = new Set(acceptance.verificationCommands.map(canonicalCommand2));
-        const receiptFound = availableEvidence.some((evidence) => evidence.success !== false && evidence.receipt?.exitCode === 0 && evidence.receipt?.timedOut !== true && expected.has(canonicalCommand2(evidence.receipt.canonicalCommand)));
+        const receiptFound = availableEvidence.some((evidence) => evidence.success !== false && evidence.receipt?.exitCode === 0 && evidence.receipt?.timedOut !== true && evidence.receipt?.cancelled !== true && expected.has(canonicalCommand2(evidence.receipt.canonicalCommand)));
         if (!receiptFound) {
           gaps.push({
             claimId: claim.id,
@@ -389148,16 +389192,10 @@ var AgenticRunner = class _AgenticRunner {
     const requirements = [];
     const evidencePredicates = [];
     if (todo.verifyCommand) {
-      requirements.push(`verifyCommand: ${todo.verifyCommand}`);
-      evidencePredicates.push({
-        schema: WORKBOARD_EVIDENCE_PREDICATE_SCHEMA,
-        id: `todo:${todo.id}:verifier`,
-        label: `successful live verifier: ${todo.verifyCommand}`,
-        tool: "shell",
-        command: todo.verifyCommand,
-        requireTrustedSuccess: true
-      });
+      requirements.push(`advisory verifyCommand: ${todo.verifyCommand}`);
     }
+    if (todo.completionJustification)
+      requirements.push(`agent assessment: ${todo.completionJustification}`);
     if (todo.declaredArtifacts?.length) {
       requirements.push(`declaredArtifacts: ${todo.declaredArtifacts.slice(0, 5).join(", ")}`);
       for (const [index, artifact2] of todo.declaredArtifacts.filter((candidate) => typeof candidate === "string").map((candidate) => this._normalizeEvidencePath(candidate)).filter(Boolean).slice(0, 5).entries()) {
@@ -389200,7 +389238,8 @@ var AgenticRunner = class _AgenticRunner {
       evidenceRequirements: requirements,
       evidencePredicates: evidencePredicates.length > 0 ? evidencePredicates : void 0,
       ownedFiles: (todo.declaredArtifacts ?? []).filter((path8) => typeof path8 === "string").map((path8) => this._normalizeEvidencePath(path8)).filter(Boolean),
-      verifierCommand: todo.verifyCommand
+      // Checklist hints do not create a mandatory mutation/compile contract.
+      verifierCommand: void 0
     };
   }
   _mirrorTodosToWorkboard(todos) {
@@ -390915,20 +390954,15 @@ ${renderTaskExecutionSnapshot(snapshot)}` } : projected;
     }
     const eligible = evidence.slice(lastInvalidatingMutation + 1).filter((entry) => entry.role === "verification" && Boolean(entry.receipt));
     const desired = [];
-    for (const todo of input.todos) {
-      if (todo.verifyCommand?.trim()) {
-        desired.push({ id: `todo-${todo.id}`, command: todo.verifyCommand });
-      }
-    }
     for (const card of input.workboard?.cards ?? []) {
-      if (!card.supersededBy && card.verifierCommand?.trim()) {
+      if (!card.supersededBy && !card.sourceTodoIds?.length && !card.id.startsWith("todo-") && card.verifierCommand?.trim()) {
         desired.push({
           id: `workboard-${card.id}`,
           command: card.verifierCommand
         });
       }
     }
-    const configuredVerifier = this.options.completionVerifyCommand?.trim() || process.env["OMNIUS_COMPLETION_VERIFY_CMD"]?.trim() || this._declaredVerifierCommand?.trim() || "";
+    const configuredVerifier = this.options.completionVerifyCommand?.trim() || process.env["OMNIUS_COMPLETION_VERIFY_CMD"]?.trim() || "";
     if (configuredVerifier) {
       desired.push({ id: "host-final-verifier", command: configuredVerifier });
     }
@@ -393969,7 +394003,7 @@ ${read.content}`;
       });
     }
   }
-  async executeToolWithInterruption(tool, name, args, onProgress, callId, assertDispatchScope) {
+  async executeToolWithInterruption(tool, name, args, onProgress, callId, assertDispatchScope, sourceCallId) {
     const progressOnly = tool.purpose === "progress";
     if (this._interruptionMode !== "off" && this._paused) {
       const shouldContinue = await this.waitIfPaused();
@@ -394225,6 +394259,15 @@ ${read.content}`;
           }
         };
       }
+      if (invocation && (name === "shell" || name === "background_run")) {
+        invocation.shellRun = {
+          callId: sourceCallId ?? callId,
+          runId: this.currentArtifactRunId(),
+          sessionId: this._sessionId,
+          taskEpoch: this._taskEpoch,
+          ownerId: this._interruptionOwnerId
+        };
+      }
       if (invocation && name === "playwright_browser" && this._hasStartedTask && this._taskEpoch > 0) {
         const goal = this._taskObjectiveContract?.goal ?? this._taskState.originalGoal;
         if (goal && goal.length <= 1e4) {
@@ -394398,7 +394441,7 @@ ${workflowStatus}` } : {}
     if (negativeScope && result.runtimeAuthored !== true && negativeScope.taskEpoch === this._taskEpoch && !this.aborted) {
       const receipt = result.executionReceipt;
       const ownedReceipt = receipt?.schema === "omnius.command-evidence-receipt.v1" && receipt.command === String(args["command"] ?? args["cmd"] ?? "") && typeof receipt.cwd === "string" && receipt.cwd.length > 0 && _pathResolve(receipt.cwd) === this.authoritativeWorkingDirectory();
-      const negative = Boolean(ownedReceipt && result.success === true && !receipt.timedOut && receipt.observation?.kind === "search_no_match");
+      const negative = Boolean(ownedReceipt && result.success === true && !receipt.timedOut && !receipt.cancelled && receipt.observation?.kind === "search_no_match");
       this._negativeObservationRecovery.observe({
         ...negativeScope,
         negative,
@@ -398196,33 +398239,15 @@ Pick the SMALLEST concrete deliverable from the spec — typically the project e
     }
     return gaps;
   }
-  /**
-   * Keep the in-memory REG-37 failure set aligned with the ordered shell
-   * history.  A completion hold can legitimately tell the model to run an
-   * exact verifier later in the same run; that later pass must clear the
-   * original hold before the truth auto-finalizer evaluates the next turn.
-   *
-   * This deliberately recomputes from post-mutation evidence rather than
-   * clearing the most recently mentioned todo.  That preserves freshness for
-   * every completed leaf and leaves independent artifact-inspection failures
-   * intact.
-   */
+  /** Retire legacy checklist-verifier holds. This does not create or refresh
+   * validation evidence; explicit acceptance and artifact checks stay separate. */
   _refreshCompletedTodoVerifierFailures(todos, toolCallLog) {
-    const gaps = this._completionVerifierGapsAfterFinalMutation(todos, toolCallLog);
-    const gapContents = new Set(gaps.map((gap) => gap.todo.content));
+    void toolCallLog;
     for (const todo of todos) {
-      const verifyCommand = typeof todo.verifyCommand === "string" ? todo.verifyCommand.trim() : "";
-      if (todo.status !== "completed" || !verifyCommand || !isExecutableVerifyCommand(verifyCommand)) {
-        continue;
-      }
-      if (gapContents.has(todo.content)) {
-        this._verifyCommandFailures.add(todo.content);
-      } else {
-        this._verifyCommandFailures.delete(todo.content);
-      }
+      this._verifyCommandFailures.delete(todo.content);
       this._refreshTodoVerificationFailure(todo.content);
     }
-    return gaps;
+    return [];
   }
   _formatFreshVerifierGapFeedback(gaps) {
     const lines = gaps.slice(0, 8).map((gap) => {
@@ -398349,12 +398374,12 @@ ${extras.join("\n")}`;
   _observeCompileVerifierResult(input) {
     const diagnostics = rankCompilerDiagnostics(parseCompilerDiagnostics(input.output));
     const verifierCommand = stripVerifierFallbackSuffix(input.verifierCommand) || input.verifierCommand.trim();
+    if (!this._compileFixLoop && diagnostics.length === 0)
+      return null;
     const previousDiagnostics = this._lastDeclaredVerifierDiagnostics;
     this._declaredVerifierCommand = verifierCommand;
     this._lastDeclaredVerifierOutput = input.output;
     this._lastDeclaredVerifierDiagnostics = diagnostics;
-    if (!this._compileFixLoop && diagnostics.length === 0)
-      return null;
     const verifierFingerprint = compileFixVerifierFingerprint(verifierCommand);
     const existingCards = this._compileFixLoop?.cards ?? [];
     const cards = diagnostics.length > 0 ? buildCompileFixCards({
@@ -399173,7 +399198,7 @@ ${modelVisible}` : modelVisible || result.error || displayOutput || "";
           receipts.unshift(legacyReceipt);
       }
       receipts.forEach((receipt2, index) => {
-        const succeeded = !!receipt2 && receipt2.exitCode === 0 && receipt2.timedOut === false && (receipt2.primaryExitCode === void 0 || receipt2.primaryExitCode === 0) && receipt2.declaredVerification?.outcome !== "fail";
+        const succeeded = !!receipt2 && receipt2.exitCode === 0 && receipt2.timedOut === false && receipt2.cancelled !== true && (receipt2.primaryExitCode === void 0 || receipt2.primaryExitCode === 0) && receipt2.declaredVerification?.outcome !== "fail";
         this._recordCompletionToolEvidenceFromResult({
           ...input,
           workOperation,
@@ -399400,7 +399425,7 @@ ${modelVisible}` : modelVisible || result.error || displayOutput || "";
     if (toolName !== "shell" || result.success !== true)
       return [];
     const receipt = result.executionReceipt;
-    if (!receipt || receipt.schema !== "omnius.command-evidence-receipt.v1" || receipt.exitCode !== 0 || receipt.timedOut || !receipt.command.trim() || !receipt.canonicalCommand.trim() || !receipt.cwd.trim()) {
+    if (!receipt || receipt.schema !== "omnius.command-evidence-receipt.v1" || receipt.exitCode !== 0 || receipt.timedOut || receipt.cancelled === true || !receipt.command.trim() || !receipt.canonicalCommand.trim() || !receipt.cwd.trim()) {
       return [];
     }
     const command = receipt.command;
@@ -401708,7 +401733,7 @@ ${contentPreview}
             toolName: input.toolName,
             fingerprint: input.fingerprint,
             turn: input.turn,
-            receipt: input.receipt && !input.receipt.timedOut && input.receipt.exitCode === 0 ? input.receipt : void 0
+            receipt: input.receipt && !input.receipt.timedOut && !input.receipt.cancelled && input.receipt.exitCode === 0 ? input.receipt : void 0
           }
         });
         for (const failure of failures) {
@@ -405229,7 +405254,7 @@ ${findingsBody}`;
     const props = this.schemaProperties(parameters);
     const required = this.schemaRequired(parameters);
     const providedKeys = Object.keys(args ?? {}).filter((key) => !key.startsWith("_"));
-    const missingRequired = required.filter((field) => args[field] === void 0 || args[field] === null);
+    const missingRequired = required.filter((field) => args[field] === void 0);
     const unknownKeys = providedKeys.filter((key) => !props.has(key));
     const aliases = this.suggestArgumentAliases(missingRequired, providedKeys, args, props);
     const corrected = this.buildCorrectedArgsPreview(args, props, aliases);
@@ -405331,7 +405356,7 @@ ${findingsBody}`;
     if (value !== null && typeof value === "object" && !Array.isArray(value)) {
       const objectValue = value;
       const required = Array.isArray(schema["required"]) ? schema["required"].filter((entry) => typeof entry === "string") : [];
-      const missing = required.filter((key) => objectValue[key] === void 0 || objectValue[key] === null);
+      const missing = required.filter((key) => objectValue[key] === void 0);
       if (missing.length > 0) {
         return `${path7} is missing required parameter(s): ${missing.join(", ")}`;
       }
@@ -405628,6 +405653,7 @@ ${findingsBody}`;
           targetPaths: this._extractToolTargetPaths(resolved?.name ?? name, args, result),
           mutatedFiles: this._isRealProjectMutation(resolved?.name ?? name, result) ? this._extractToolTargetPaths(resolved?.name ?? name, args, result) : [],
           executionReceipt: result.executionReceipt,
+          shellRunId: result.shellRunId ?? result.executionReceipt?.shellRunId,
           webFetchReceipt: this._webFetchEventReceipt({ toolName: name, callId, args, result, scope: evidenceScope }),
           pageContentReceipt: this._pageContentEventReceipt({ toolName: name, callId, args, result, scope: evidenceScope }),
           content: this.toolResultEventContent(name, result.output || result.error || ""),
@@ -409989,7 +410015,6 @@ ${text13}`;
     if (candidates.length !== 1)
       return null;
     const card = candidates[0];
-    const verifier = card.verifierCommand?.trim();
     const workItem = {
       cardId: card.id,
       title: card.title,
@@ -409997,7 +410022,7 @@ ${text13}`;
       todoIds: card.sourceTodoIds ?? [active.id],
       ownedFiles: card.ownedFiles ?? active.declaredArtifacts ?? [],
       expectedArtifacts: active.declaredArtifacts ?? card.ownedFiles ?? [],
-      verifierCommand: verifier || active.verifyCommand,
+      verifierCommand: void 0,
       evidenceRequirements: card.evidenceRequirements
     };
     const evidence = [
@@ -411762,6 +411787,7 @@ Respond with the assessment and take the selected evidence-backed action.`;
             parentId: t.parentId,
             blocker: t.blocker,
             verifyCommand: t.verifyCommand,
+            completionJustification: t.completionJustification,
             declaredArtifacts: t.declaredArtifacts
           })));
           const lines = verification.downgrades.slice(0, 6).map((d) => `- "${d.content.slice(0, 90)}": ${d.from} -> ${d.to} (${d.evidence})`);
@@ -412609,7 +412635,7 @@ ${dynamicProjectContext}`
           `Unresolved verification item(s) introduced this run:`,
           ...failures.map((item) => `  - ${item}`),
           ``,
-          `Run the missing verifyCommand exactly, repair the todo evidence, or mark the todo blocked. Do not call task_complete until the verification state is clean. (Pre-existing failures present at task start are ignored.)`
+          `Inspect the missing declared artifact or describe the remaining gap accurately. Optional verifyCommand hints do not require another shell run. (Pre-existing failures present at task start are ignored.)`
         ].join("\n");
         lastCompletionGateReason = `${newFailures.length} unresolved todo verification failure(s) introduced this run`;
         lastCompletionGateFeedback = feedback;
@@ -412977,14 +413003,6 @@ ${lastCompletionGateFeedback.trim().slice(0, 4e3)}` : ""
           ].join("\n")
         };
       }
-      const verifierGaps = this._completionVerifierGapsAfterFinalMutation(this.readSessionTodos() || [], toolCallLog).filter((gap) => gap.lastMutationTurn >= 0);
-      if (verifierGaps.length > 0) {
-        return {
-          code: "post_mutation_verification",
-          reason: `${verifierGaps.length} completed todo verifier(s) have no successful receipt after the final relevant mutation`,
-          feedback: this._formatFreshVerifierGapFeedback(verifierGaps)
-        };
-      }
       const compileFixLoopWasMutated = this._compileFixLoop?.lastMutationTurn !== null && this._compileFixLoop?.lastMutationTurn !== void 0;
       const compileFeedback = compileFixLoopWasMutated ? this._compileFixLoopCompletionBlocker(turn) : null;
       if (compileFeedback) {
@@ -414169,7 +414187,7 @@ ${cachedResult}`,
           if (!tool.inputSchema && !validationError && tool.parameters && typeof tool.parameters === "object") {
             const required = tool.parameters.required;
             if (Array.isArray(required)) {
-              const missing = required.filter((field) => tc.arguments[field] === void 0 || tc.arguments[field] === null);
+              const missing = required.filter((field) => tc.arguments[field] === void 0);
               if (missing.length > 0) {
                 validationError = `Missing required parameter(s): ${missing.join(", ")}`;
               }
@@ -414362,10 +414380,11 @@ ${cachedResult}`,
                         this.emit({
                           type: "status",
                           toolName: tc.name,
+                          actionId: tc.id,
                           content: progress,
                           timestamp: (/* @__PURE__ */ new Date()).toISOString()
                         });
-                      }, evidenceCallId);
+                      }, evidenceCallId, void 0, tc.id);
                       result = await this._guardMaterializedFileReadResult({
                         callId: tc.id,
                         toolName: resolvedTool?.name ?? tc.name,
@@ -415083,9 +415102,10 @@ ${diagnostic}`;
           } catch {
           }
         }
+        const processWasStopped = result.executionReceipt?.cancelled === true;
         {
           const _cmd = typeof tc.arguments?.["command"] === "string" ? tc.arguments["command"] : typeof tc.arguments?.["cmd"] === "string" ? tc.arguments["cmd"] : "";
-          if (_cmd && tc.name === "shell") {
+          if (_cmd && tc.name === "shell" && !processWasStopped) {
             if (!result.success) {
               this._resolutionMemory.recordFailure({
                 tool: tc.name,
@@ -415112,7 +415132,7 @@ ${diagnostic}`;
           }
         }
         let failureWorkboardCardId = null;
-        if (!result.success && !completionReviewIsPending && (result.error || result.output || result.llmContent)) {
+        if (!result.success && !processWasStopped && !completionReviewIsPending && (result.error || result.output || result.llmContent)) {
           const errorText = (result.error || result.output || result.llmContent || "").toString();
           const diagnosis = this._diagnoseToolFailure(tc.name, tc.arguments, result);
           const sig = diagnosis.patternKey;
@@ -415266,6 +415286,7 @@ ${diagnostic}`;
                   parentId: t.parentId,
                   blocker: t.blocker,
                   verifyCommand: t.verifyCommand,
+                  completionJustification: t.completionJustification,
                   declaredArtifacts: t.declaredArtifacts
                 })));
                 _todosNow = this.readSessionTodos() || truth.todos;
@@ -415288,19 +415309,19 @@ ${diagnostic}`;
                 });
               }
               this._mirrorTodosToWorkboard(_todosNow);
-              if (!this._newFieldNudgeFired && !this._isTodoWriteNoopResult(result)) {
+              if (!this._newFieldNudgeFired && _todosNow.some((t) => t.status !== "completed") && !this._isTodoWriteNoopResult(result)) {
                 this._todoWritesObservedForNudge++;
-                const _anyFieldUsed = _todosNow.some((t) => typeof t.verifyCommand === "string" || Array.isArray(t.declaredArtifacts));
+                const _anyFieldUsed = _todosNow.some((t) => typeof t.verifyCommand === "string" || Boolean(t.completionJustification?.trim()) || Array.isArray(t.declaredArtifacts));
                 if (this._todoWritesObservedForNudge >= 2 && !_anyFieldUsed) {
                   this._newFieldNudgeFired = true;
                   pushSoftInjection("system", [
                     `[NUDGE — REG-37e: you have emitted multiple todo_writes without using verifyCommand or declaredArtifacts.]`,
                     ``,
-                    `These two fields turn self-reported completion into VERIFIED completion. The orchestrator auto-checks them when you mark a todo "completed":`,
+                    `Optional verification metadata can help explain completion. A declared command is not a passing result; use actual receipts or a stated completionJustification:`,
                     `  - verifyCommand: a shell invocation that proves the work passes (test runner, build command, file existence check, etc.)`,
                     `  - declaredArtifacts: list of file paths this todo produces`,
                     ``,
-                    `Without these, your "completed" claim is a self-report. With them, it's checked against reality. The very next todo you write where "done" has an objective check should include one or both fields.`,
+                    `Completion may be a justified agent assessment. Record the basis and limitations; do not add commands or repeat work solely to fill optional checklist metadata.`,
                     ``,
                     `Worked example shape (substitute commands native to your stack):`,
                     `  {"id":"pX","content":"Implement the cache module","status":"in_progress","verifyCommand":"<your test command>","declaredArtifacts":["src/lib/cache.ts","tests/unit/cache.test.ts"]}`
@@ -415310,40 +415331,8 @@ ${diagnostic}`;
               for (const _t of _todosNow) {
                 if (_t.status !== "completed")
                   continue;
-                const _vc = _t.verifyCommand;
-                if (_vc && typeof _vc === "string" && isExecutableVerifyCommand(_vc) && !this._verifyHintInjectedThisTurn.has(_t.content)) {
-                  const _history = toolCallLog.slice(-15).filter((c) => c.name === "shell").map((c) => ({
-                    command: this._shellCommandFromArgsKey(c.argsKey),
-                    success: c.success === true,
-                    receipt: c.executionReceipt
-                  }));
-                  const _verified = verifyCommandSatisfiedByShellHistory(_vc, _history);
-                  if (_verified) {
-                    this._verifyCommandFailures.delete(_t.content);
-                    this._refreshTodoVerificationFailure(_t.content);
-                  } else {
-                    this._verifyCommandFailures.add(_t.content);
-                    this._refreshTodoVerificationFailure(_t.content);
-                    this._verifyHintInjectedThisTurn.add(_t.content);
-                    messages2.push({
-                      role: "system",
-                      content: [
-                        `[VERIFICATION REQUIRED — REG-37]`,
-                        ``,
-                        `You marked the todo "${_t.content.slice(0, 100)}" as completed, but its verifyCommand has not been run successfully in your recent shell history.`,
-                        ``,
-                        `verifyCommand: \`${_vc}\``,
-                        ``,
-                        `Run this command BEFORE the completion is accepted as real progress. If it succeeds, the completion is verified. If it fails, revert the todo to in_progress and fix the underlying issue.`,
-                        ``,
-                        `Per the MAST failure taxonomy (Cemri et al. NeurIPS 2025), 21.3% of multi-agent failures are "task verification & termination" — claiming completion without running the validation that proves it. Don't be one of them.`
-                      ].join("\n")
-                    });
-                  }
-                } else {
-                  this._verifyCommandFailures.delete(_t.content);
-                  this._refreshTodoVerificationFailure(_t.content);
-                }
+                this._verifyCommandFailures.delete(_t.content);
+                this._refreshTodoVerificationFailure(_t.content);
                 if (!this._artifactInspectionDoneThisTurn.has(_t.content)) {
                   this._artifactInspectionDoneThisTurn.add(_t.content);
                   const _writeMap = /* @__PURE__ */ new Map();
@@ -415572,7 +415561,7 @@ ${diagnostic}`;
             }
           }
         }
-        if (!result.success && result.runtimeAuthored !== true && !this._isPdfTerminalFeedback(result)) {
+        if (!result.success && !processWasStopped && result.runtimeAuthored !== true && !this._isPdfTerminalFeedback(result)) {
           const failureTargets = this._extractToolTargetPaths(tc.name, tc.arguments, result);
           const failureDiagnosis = result.failureDiagnosis ?? this._diagnoseToolFailure(tc.name, tc.arguments, result);
           this._recentFailures.push({
@@ -415720,7 +415709,7 @@ ${bookkeepingGuidance}`;
 
 ${bookkeepingGuidance}` : bookkeepingGuidance;
         }
-        if (!result.success && !completionReviewIsPending && this.options.modelTier === "small") {
+        if (!result.success && !processWasStopped && !completionReviewIsPending && this.options.modelTier === "small") {
           const recovery = this.buildRecoveryGuidance(tc.name, result.error ?? "", tc.arguments);
           if (recovery)
             output += "\n\n" + recovery;
@@ -415742,7 +415731,7 @@ ${bookkeepingGuidance}` : bookkeepingGuidance;
         }
         if (result.success) {
           delete result.failureDiagnosis;
-        } else if (!completionReviewIsPending && !this._isPdfTerminalFeedback(result)) {
+        } else if (!processWasStopped && !completionReviewIsPending && !this._isPdfTerminalFeedback(result)) {
           this._diagnoseToolFailure(tc.name, tc.arguments, result);
           const staleHashRetryGuidance = this._buildStaleHashRetryGuidance(tc.name, tc.arguments, result);
           if (staleHashRetryGuidance) {
@@ -415908,6 +415897,7 @@ Evidence: ${evidencePreview}`.slice(0, 500);
           targetPaths: toolTargetPaths,
           mutatedFiles: realMutationPaths,
           executionReceipt: result.executionReceipt,
+          shellRunId: result.shellRunId ?? result.executionReceipt?.shellRunId,
           synthetic: result.runtimeAuthored === true,
           blocked: result.runtimeAuthored === true && result.success === false,
           ...this._objectiveFrontier ? { objectiveFrontier: this._objectiveFrontier } : {},
@@ -416106,7 +416096,7 @@ Evidence: ${evidencePreview}`.slice(0, 500);
 ${structuralGuardGuidance}` : structuralGuardGuidance;
           }
         }
-        if (tc.name === "shell" && result.runtimeAuthored !== true) {
+        if (tc.name === "shell" && !processWasStopped && result.runtimeAuthored !== true) {
           const shellCommand = String(tc.arguments?.["command"] ?? tc.arguments?.["cmd"] ?? "");
           const compileOutput = this._expandTriagedPayloadForDiagnostics([result.output, result.llmContent, result.error].filter((part) => typeof part === "string" && part.length > 0).join("\n"));
           const verifierCommand = this._declaredVerifierCommandForShell(shellCommand, compileOutput, result.success === true, result.executionReceipt);
@@ -416397,7 +416387,7 @@ Do NOT retry ${tc.name} with similar arguments and do NOT use file_write to esca
           if (isModify && turnTier === "small") {
             const modCount = this._taskState.modifiedFiles.size;
             if (modCount >= 2 && modCount % 2 === 0) {
-              this.enqueueRuntimeGuidance(`[Test reminder] You've modified ${modCount} files. Run relevant tests NOW to verify: shell(command="npm test") or the project's test command. Fix any failures before continuing.`);
+              this.enqueueRuntimeGuidance(`[Test reminder] You've modified ${modCount} files. Run relevant tests NOW to verify: shell(command="npm test", timeout=300000) or the project's test command with an explicit deadline (or null). Fix any failures before continuing.`);
             }
           }
         }
@@ -422116,6 +422106,9 @@ ${marker}` : marker);
       if (typeof args["verifyCommand"] === "string" && args["verifyCommand"].trim()) {
         todo["verifyCommand"] = args["verifyCommand"].trim();
       }
+      if (typeof args["completionJustification"] === "string" && args["completionJustification"].trim()) {
+        todo["completionJustification"] = args["completionJustification"].trim();
+      }
       if (typeof args["blocker"] === "string" && args["blocker"].trim()) {
         todo["blocker"] = args["blocker"].trim();
       }
@@ -425376,7 +425369,7 @@ Do NOT rewrite the entire function. Patch the specific fault, then re-run tests.
       const taskGoal = (this._taskState.goal || "").toLowerCase();
       const isDebugging = /bug|fix|debug|broken|wrong|fail|error|issue|refactor|extract/i.test(taskGoal);
       if (isDebugging) {
-        return `You just read "${lastFilePath}". Before changing anything, WRITE A QUICK TEST to observe the current behavior: ${this._promptToolAvailable("shell") ? 'shell(command="node -e …")' : this._promptToolAvailable("repl_exec") ? "repl_exec" : "an available inspection tool"}. See what actually happens — don't guess. Then fix based on what you observed.`;
+        return `You just read "${lastFilePath}". Before changing anything, WRITE A QUICK TEST to observe the current behavior: ${this._promptToolAvailable("shell") ? 'shell(command="node -e …", timeout=30000)' : this._promptToolAvailable("repl_exec") ? "repl_exec" : "an available inspection tool"}. See what actually happens — don't guess. Then fix based on what you observed.`;
       }
       return `You just read "${lastFilePath}". Now make the needed changes with file_edit, or if exploring, decide what to modify.`;
     }
@@ -432589,6 +432582,7 @@ function mapExecutionToolResult(result) {
     // uses this receipt for completion evidence; reconstructing it from a
     // truncated/streamed transcript would be unsound.
     executionReceipt: result.executionReceipt,
+    shellRunId: result.shellRunId ?? result.executionReceipt?.shellRunId,
     executionReceipts: result.executionReceipts,
     documentExtraction: result.documentExtraction,
     runtimeSession: result.runtimeSession,