fix: bind AgentRun lane policy and bounded output

project-management/PJ2026-01/specs/PJ2026-010202-runtime-assembly.md

@@ -37,6 +37,7 @@ Runtime装配负责把一次 AgentRun 执行需要的镜像、provider profile
 - provider credential、Git resource credential、tool credential 和 short-lived execution context 的 SecretRef/projection/redaction 归属。
 - Git-only `ResourceBundleRef.kind="gitbundle"`、`bundles[]`、`promptRefs[]`、`requiredSkills[]`、workspace materialization 和实际 commit 摘要。
 - AipodSpec YAML、`imageRef`、env image reuse、Artificer 默认装配和 `--aipod` render 入口。
+- AipodSpec render 对目标 node/lane 的 backendProfile、provider credential SecretRef、tool credential SecretRef、workspaceRef 和 bundle 绑定。
 - SessionRef 到 PVC 直接挂载、`CODEX_HOME`、profile runtime home 和 Secret projection 分离。
 
 ### 2.3 范围外
@@ -57,6 +58,7 @@ Runtime装配负责把一次 AgentRun 执行需要的镜像、provider profile
 | SessionRef | 描述 backend 会话、thread、PVC 和续跑状态的运行时要素。 |
 | ResourceBundleRef | 描述 Git-only 资源包、工具、skill、prompt 和实际物化 commit 的运行时要素。 |
 | AipodSpec | 声明式 agent 装配 YAML，用于集中描述 imageRef、backendProfile、executionPolicy、resourceBundleRef 和 payloadDefaults。 |
+| AipodSpec lane binding | AipodSpec render 时将声明式装配与目标 node/lane 的 YAML SecretRef、workspaceRef、bundle 和 provider profile 绑定的过程。 |
 | toolCredentials | 运行时工具授权的 SecretRef 装配项，例如 GitHub PR token、GitHub SSH key 或 UniDesk SSH passthrough token。 |
 | transientEnv | 单次 runner job 的短期运行上下文，通过短期 Secret 投影，不作为 run durable fact。 |
 
@@ -105,6 +107,8 @@ Runtime装配应通过 SecretRef 和 projection intent 分发 provider credentia
 
 输出只能包含 SecretRef 名称、key 名、projection kind、presence、hash/fingerprint、ownerReference 状态和 `valuesPrinted=false`。provider API key、Codex `auth.json`、`config.toml`、DSN、token、SSH private key 和 transientEnv value 都不得以明文形式出现在响应、dry-run、event 或日志中。
 
+RuntimeAssembly 或 AipodSpec render 缺少目标 lane 所需的 provider credential SecretRef、tool credential SecretRef 或 sourceRef 时，应报告为 YAML/AipodSpec binding 缺口。修复路径是补齐 YAML/sourceRef、受控 Secret 同步和重新 render，不允许通过手工创建 legacy Secret、复制其他 lane Secret 或 patch runtime namespace 规避装配缺陷。
+
 ### 6.3 AR-RUNTIME-REQ-003 GitBundle 资源物化
 
 | 编号 | 短名 | 主责模块 | 关联模块 |
@@ -136,3 +140,5 @@ SessionRef 是 Web、CLI、Queue 和 HWLAB接入共享的会话连续性 authori
 Runtime装配应提供 AipodSpec YAML，使 imageRef、backendProfile、executionPolicy、resourceBundleRef、model 和 payloadDefaults 可以从声明式规格进入 Queue submit 或 Session send。
 
 AipodSpec 只声明装配意图，不保存 API key、SSH private key、token、`auth.json`、`config.toml` 或其他 Secret 明文。`imageRef` 必须指向可追溯的 env image Dockerfile source 和完整 commit，最终 runner job 只能使用解析后的 digest-pinned image 或结构化报告 build-required。
+
+当 Queue submit、Session send 或 CLI `--aipod` render 已指定 node/lane 时，AipodSpec 必须使用该目标 lane 的 YAML 事实绑定 backendProfile、provider credential SecretRef、tool credential SecretRef、workspaceRef 和 resource bundle。除非 YAML 明确声明继承，AipodSpec 不得从其他 lane 或全局默认 lane 隐式继承 provider、tool credential、workspace 或 namespace。

project-management/PJ2026-01/specs/PJ2026-010203-queue-session.md

@@ -37,6 +37,7 @@
 - Session API/CLI 的输出、trace、命令流、debug/audit 详情、read 状态和会话控制。
 - Queue task 到 Core run/command/runner job/session 的引用关系和 sessionPath 输出。
 - `sessions send`、session continuation、unread/default/all 状态和 terminal projection。
+- `send session` follow-up 对 node/lane、backendProfile、providerId、workspaceRef、executionPolicy 和 provider credential SecretRef 的策略解析。
 - Queue task、Session、run 和 command 取消入口的用户语义、级联边界、同 session 续跑和 canceled terminal projection。
 - 自动 scheduler 的 deferred 边界、future pending scan、capacity selection、runner assignment 和 stale lease recovery 方向。
 
@@ -56,6 +57,7 @@
 | Queue task | 用户或系统提交到 AgentRun Queue 的任务资源，可 dispatch 为 Core run/command。 |
 | attempt | Queue task 的一次执行尝试，引用 runId、commandId、runnerJobId 和 sessionId。 |
 | Session | 面向用户的执行会话控制面，承载 output、trace、read、send、cancel 和 terminal projection。 |
+| session policy | Session follow-up 创建下一次 run/command 时使用的 backendProfile、providerId、workspaceRef、executionPolicy 和 provider credential SecretRef 策略。 |
 | sessionPath | Queue task 返回的 Session API 路径，用于读取输出和 trace。 |
 | read cursor | 按 readerId 记录的已读水位，用于默认列表只显示 running 或 unread session/task。 |
 | commander | Queue 侧的聚合视图，用于展示队列状态、最新 attempt 和下一步操作摘要。 |
@@ -69,7 +71,7 @@
 | 边界项 | 内容 |
 | --- | --- |
 | 外部使用者 | UniDesk 用户、HWLAB Agent 入口、CLI 操作者、Artificer/Aipod、AgentRun 核心和平台运维。 |
-| 外部输入 | Queue task JSON、dispatch body、sessionId、prompt、run base、readerId、backendProfile、resourceBundleRef、cancel/read/send 控制请求。 |
+| 外部输入 | Queue task JSON、dispatch body、sessionId、prompt、run base、readerId、node/lane、backendProfile、providerId、workspaceRef、resourceBundleRef、cancel/read/send 控制请求。 |
 | 受控资源 | Queue task、attempt、summary、stats、read cursor、Session projection、session trace/output 和 scheduler deferred 状态。 |
 | 外部输出 | taskId、attemptId、sessionId、sessionPath、queue stats、commander summary、unread/default/all 列表、output/trace cursor 和控制结果。 |
 | 用户接口 | AgentRun REST Queue/Session API、`./scripts/agentrun queue ...`、`./scripts/agentrun sessions ...`。 |
@@ -107,6 +109,10 @@ AgentRun Queue 直接吸收 UniDesk Code Queue 的新任务入口，不做 adapt
 
 Queue task 详情只能返回 `sessionPath` 和有界摘要，不代理完整输出或 trace。输出、trace 和会话控制必须从 Session API/CLI 进入，避免 Queue 列表视图反推 Core event stream 或展开执行细节。
 
+`send session/<sessionId>` 的短命令形态、显式 JSON body 形态和 `--aipod` 形态都必须使用目标 node/lane 的 YAML 事实解析 session policy。CLI 或 API 已显式给出 `node`/`lane` 时，backendProfile、providerId、workspaceRef、executionPolicy 和 provider credential SecretRef 不得回退到全局默认 lane；未显式给出目标时，输出必须披露使用的 YAML default 来源。
+
+`explain session-policy` 是 `send session` 的同源预检入口，必须展示与实际 follow-up 创建相同的目标 node/lane、backendProfile、providerId、workspaceRef、executionPolicy、provider credential SecretRef 和脱敏来源摘要。若解释输出与目标 lane 不一致，应视为策略解析或渲染缺陷，不得通过复制其他 lane Secret、手工补 legacy Secret 或修改 runtime namespace 绕过。
+
 ### 6.3 AR-QUEUE-REQ-003 Queue/Session/Core 分层
 
 | 编号 | 短名 | 主责模块 | 关联模块 |

project-management/PJ2026-01/specs/PJ2026-010603-yaml-first-ops.md

@@ -41,7 +41,7 @@ YAML运维负责 HWLAB/UniDesk 自有平台配置的真相源、解析、渲染
 - Sub2API、Codex pool、AgentRun control-plane、session policy 和平台基础设施配置的受控 CLI 读取、解释、计划和下发。
 - AgentRun lane 的 runner retention、idle timeout、egress proxy 和 cancel lifecycle policy 等运行策略配置读取、解释和下发。
 - FRP、Caddy、public URL、public health、Kubernetes Secret 和平台资源渲染所需的配置投递边界。
-- 可复用 ops primitive，包括 YAML path 捕获、字段解析、fingerprint、摘要输出、Secret 引用和命令输出约束。
+- 可复用 ops primitive，包括 YAML path 捕获、字段解析、fingerprint、摘要输出、Secret 引用、有界默认输出和命令输出约束。
 
 ### 2.3 范围外
 
@@ -66,6 +66,7 @@ YAML运维负责 HWLAB/UniDesk 自有平台配置的真相源、解析、渲染
 | publicExposure | YAML 中描述 FRP、Caddy、domain、TLS、public URL 和 health 目标的公开入口声明。 |
 | ops primitive | 平台运维 CLI 共享的底层能力，例如字段解析、fingerprint、Secret 引用、摘要输出和 YAML path 捕获。 |
 | 配置解释输出 | CLI 将 YAML 解析后的默认值、来源和目标以非敏感摘要展示给操作人员的输出。 |
+| 有界默认输出 | CLI 在未显式请求机器输出或完整输出时返回的短摘要、表格和 drill-down 命令集合。 |
 
 ## 4. 系统边界和接口
 
@@ -76,7 +77,7 @@ YAML运维负责 HWLAB/UniDesk 自有平台配置的真相源、解析、渲染
 | 外部使用者 | 平台管理员、发布操作人员、Agent编排维护者、Sub2API/Codex pool 维护者和自动化运维任务。 |
 | 外部输入 | YAML 配置文件、target/lane/node/service 声明、Secret sourceRef、providerCredential、publicExposure、运行目标选择和配置解释请求。 |
 | 受控资源 | 配置解析器、公共 ops helper、Sub2API/Codex pool 配置、AgentRun lane 配置、Kubernetes Secret 声明、FRP/Caddy 渲染、public health 和 CLI 输出摘要。 |
-| 外部输出 | 配置解释结果、计划摘要、同步摘要、Secret presence/fingerprint、publicExposure 渲染结果、运行目标解析结果和错误诊断。 |
+| 外部输出 | 配置解释结果、计划摘要、同步摘要、Secret presence/fingerprint、publicExposure 渲染结果、运行目标解析结果、有界默认输出和错误诊断。 |
 | 用户接口 | `bun scripts/cli.ts platform-infra ...`、`bun scripts/cli.ts agentrun ...`、受控 Secret 同步命令、配置解释命令和平台状态查询命令。 |
 | 系统边界 | YAML运维负责让配置从 YAML 可解释、可校验并受控进入运行面；不定义业务策略数值、不替代 Agent/用户/客户端语义，也不把运行面观测反向变成本地配置真相。 |
 
@@ -115,6 +116,8 @@ YAML运维应提供 target、lane、node、service、namespace 和默认目标
 
 当 issue、PR 或 CLI 已经明确 node/lane 时，YAML 解析只负责校验和解释该目标，不得用全局默认覆盖用户选择。没有明确目标时，CLI 可以读取 YAML default，但输出必须说明 default 来源、目标 node/lane、namespace、source truth 和公开入口，避免把历史 G14/v0.2、D601 legacy 或旧 `dev/prod` 口径误用为当前运行面。
 
+AgentRun `send session`、`explain session-policy`、resource primitives 和 AipodSpec render 都属于 node/lane 敏感入口。命令行或请求体已明确 node/lane 时，这些入口必须使用目标 lane 的 YAML 事实解析 backendProfile、providerId、workspaceRef、executionPolicy、provider credential SecretRef、tool credential SecretRef、namespace 和 source truth；不得在解释、dry-run、短命令或 JSON body 路径中回退到 control-plane default 或历史全局 sessionPolicy。
+
 ### 6.3 OPS-YAML-REQ-003 Secret 绑定与敏感输出
 
 | 编号 | 短名 | 主责模块 | 关联模块 |
@@ -125,6 +128,8 @@ YAML运维应通过 sourceRef、targetKey、providerCredential 和 manual bindin
 
 密钥相关输出只能展示对象名、key 名、sourceRef、presence、fingerprint、字节数和执行摘要，不得打印 base64 payload、解码值、完整 DSN、API key 或可复制凭据。运行面只能作为 presence 和 health 观测对象，不能被用来反推本地配置真相。
 
+目标 lane 缺少 providerCredential、tool credential、targetKey 或 sourceRef 时，应作为 YAML 配置或 AipodSpec binding 缺口暴露。受控 CLI 可以显示缺失对象名、key 名、YAML path、presence=false、fingerprint 缺失和 `valuesPrinted=false`，但不得用手工 Secret、复制其他 lane 凭据、读取运行面 Secret value 或 patch namespace 的方式修复配置缺口。
+
 ### 6.4 OPS-YAML-REQ-004 公开入口渲染
 
 | 编号 | 短名 | 主责模块 | 关联模块 |
@@ -149,6 +154,8 @@ YAML运维应为 AgentRun control-plane default、client sessionPolicy、lane se
 
 cancel lifecycle policy 至少应能声明取消信号投递方式、runner graceful abort、kill escalation、stale heartbeat fencing window、late write fencing 和默认事件阶段输出开关。CLI 只校验字段结构、类型、必填项和可渲染性；具体窗口、超时和开关值由 YAML 承载，不在代码或 SPEC 中写成第二真相。
 
+sessionPolicy 的解释、Session follow-up 的 run body、AipodSpec render 结果和 resource primitive 的 dry-run/JSON 形态必须来自同一套 YAML 目标解析。非默认 lane 的 backendProfile、providerId、workspaceRef、executionPolicy 和 provider credential SecretRef 不得被全局 default lane 覆盖；输出中应明确目标 lane 与 policy 来源，使操作者能在创建 run/command 前发现 lane 绑定错误。
+
 本需求只约束执行策略如何作为平台配置进入运行面。Agent run、command、session 状态机、任务恢复、取消语义和 provider 业务语义由 Agent编排负责，用户身份和 API key 约束由用户管理负责。
 
 ### 6.6 OPS-YAML-REQ-006 公共 ops primitive
@@ -160,3 +167,5 @@ cancel lifecycle policy 至少应能声明取消信号投递方式、runner grac
 YAML运维应沉淀公共 ops primitive，使字段解析、YAML path 捕获、fingerprint、Secret 引用、摘要输出和配置错误说明在 Sub2API、AgentRun 和后续平台 CLI 中复用。
 
 公共 ops primitive 的职责是减少重复硬编码和输出口径漂移。领域 CLI 仍负责自身命令形态和服务语义，但不得复制一套不同的敏感输出、fingerprint、字段解析或 YAML 缺失处理规则。
+
+受控 CLI 的默认输出必须是有界摘要、关键字段表格和下一步 drill-down 命令；`/tmp/unidesk-cli-output` dump 只能作为异常兜底，不能成为正常交互路径。完整 JSON、raw payload、完整 task/result/log、长 plan 和机器消费输出只能在显式 `--full`、`--raw`、`-o json` 或等价机器消费参数下返回，并仍须遵守 Secret 脱敏规则。

scripts/src/agentrun.ts

@@ -1,5 +1,5 @@
-// SPEC: PJ2026-01020108 cancel lifecycle + PJ2026-01020305 cancel control + PJ2026-01060305 AgentRun execution policy draft-2026-06-25-p0.
-// Exposes AgentRun cancel lifecycle policy and dry-run visibility in the UniDesk CLI.
+// SPEC: PJ2026-01020108 cancel lifecycle + PJ2026-01020205 AipodSpec binding + PJ2026-01020302 session policy + PJ2026-01020305 cancel control + PJ2026-01060305/06 YAML execution policy and bounded output draft-2026-06-25-p0.
+// Exposes AgentRun lane-scoped policy, AipodSpec SecretRef binding, cancel lifecycle, and bounded default output in the UniDesk CLI.
 import { chmodSync, copyFileSync, existsSync, readFileSync, statSync, writeFileSync } from "node:fs";
 import { join } from "node:path";
 import { spawnSync } from "node:child_process";
@@ -97,7 +97,11 @@ export async function runAgentRunCommand(config: UniDeskConfig | null, args: str
   }
   if (config === null) throw new Error("agentrun control-plane and git-mirror commands require UniDesk config");
   if (group === "control-plane") {
-    if (action === "plan") return await controlPlanePlan(config, parseStatusOptions(actionArgs));
+    if (action === "plan") {
+      const options = parseStatusOptions(actionArgs);
+      const result = await controlPlanePlan(config, options);
+      return options.full || options.raw ? result : renderAgentRunControlPlanePlanSummary(result);
+    }
     if (action === "apply") return await controlPlaneApply(config, parseLaneConfirmOptions(actionArgs));
     if (action === "status") {
       const options = parseStatusOptions(actionArgs);
@@ -108,8 +112,16 @@ export async function runAgentRunCommand(config: UniDeskConfig | null, args: str
     if (action === "restart") return await restartYamlLane(config, parseLaneConfirmOptions(actionArgs));
     if (action === "expose") return await exposeAgentRun(config, parseConfirmOptions(actionArgs));
     if (action === "trigger-current") return await triggerCurrent(config, parseTriggerOptions(actionArgs));
-    if (action === "refresh") return await refresh(config, parseRefreshOptions(actionArgs));
-    if (action === "cleanup-runners") return await cleanupRunners(config, parseCleanupRunnersOptions(actionArgs));
+    if (action === "refresh") {
+      const options = parseRefreshOptions(actionArgs);
+      const result = await refresh(config, options);
+      return options.full || options.raw ? result : renderAgentRunControlPlaneActionSummary(result, "AGENTRUN CONTROL-PLANE REFRESH");
+    }
+    if (action === "cleanup-runners") {
+      const options = parseCleanupRunnersOptions(actionArgs);
+      const result = await cleanupRunners(config, options);
+      return options.full || options.raw ? result : renderAgentRunControlPlaneActionSummary(result, "AGENTRUN RUNNER CLEANUP");
+    }
     if (action === "cleanup-runs") return await cleanupRuns(config, parseCleanupRunsOptions(actionArgs));
     if (action === "cleanup-released-pvs") return await cleanupReleasedPvs(config, parseCleanupReleasedPvOptions(actionArgs));
   }
@@ -309,13 +321,13 @@ function agentRunGetKindHelp(kindRaw: string): string {
 
 async function runAgentRunResourceCommand(config: UniDeskConfig | null, verb: AgentRunResourceVerb, action: string | undefined, actionArgs: string[], canonicalArgs: string[]): Promise<RenderedCliResult> {
   if (isHelpArg(action) || actionArgs.some(isHelpArg)) return renderAgentRunHelp(canonicalArgs);
-  if (verb === "explain") return renderedCliResult(true, "agentrun explain", agentRunExplain(action ?? "task"));
   const resourceArgs = action === undefined ? actionArgs : [action, ...actionArgs];
   const options = parseResourceOptions(resourceArgs);
   const bridgeActionArgs = stripAgentRunResourceWrapperArgs(actionArgs);
   const command = `agentrun ${canonicalArgs.join(" ")}`.trim();
   try {
     return await withAgentRunRestTarget(resolveAgentRunRestTarget(config, options), async () => {
+      if (verb === "explain") return renderedCliResult(true, command, agentRunExplain(action ?? "task", resourceArgs, options));
       if (verb === "get") return await resourceGet(config, command, action, bridgeActionArgs, options);
       if (verb === "describe") return await resourceDescribe(config, command, action, bridgeActionArgs, options);
       if (verb === "events") return await resourceEvents(config, command, action, bridgeActionArgs, options);
@@ -501,7 +513,10 @@ async function resourceDescribe(config: UniDeskConfig | null, command: string, a
     const data = record(innerData(result));
     const task = unwrapTaskDetail(data);
     if (options.raw) return renderMachine(command, result, "json", result.ok !== false);
-    if (options.output === "json" || options.output === "yaml") return renderMachine(command, { kind: ref.kind, name: ref.name, resource: task }, options.output, result.ok !== false);
+    if (options.output === "json" || options.output === "yaml") {
+      const payload = options.full ? { kind: ref.kind, name: ref.name, resource: task } : compactTaskDescriptionPayload(ref, task);
+      return renderMachine(command, payload, options.output, result.ok !== false);
+    }
     return renderedCliResult(result.ok !== false, command, renderTaskDescription(task, options));
   }
   if (ref.kind === "run") {
@@ -755,7 +770,7 @@ async function resourceCreate(config: UniDeskConfig | null, command: string, act
   if (kind !== "task") throw new Error("create currently supports: create task");
   const submitArgs = ["submit", ...stripLeadingResource(args, "task")];
   const result = await runAgentRunRestCommand(config, "queue", submitArgs);
-  return renderMutationSummary(command, result, options, "Task create submitted", options.dryRun ? [rerunWithoutDryRun(command)] : undefined);
+  return renderMutationSummary(command, result, options, options.dryRun ? "Task create plan" : "Task create submitted", options.dryRun ? [rerunWithoutDryRun(command)] : undefined);
 }
 
 async function resourceApply(config: UniDeskConfig | null, command: string, args: string[], options: AgentRunResourceOptions): Promise<RenderedCliResult> {
@@ -1515,6 +1530,46 @@ function renderTaskDescription(task: Record<string, unknown>, options: AgentRunR
   return lines.join("\n");
 }
 
+function compactTaskDescriptionPayload(ref: AgentRunResourceRef, task: Record<string, unknown>): Record<string, unknown> {
+  const attempt = record(task.latestAttempt);
+  const taskId = stringOrNull(task.id) ?? ref.name;
+  const runId = stringOrNull(attempt.runId);
+  const commandId = stringOrNull(attempt.commandId);
+  const runnerJobId = stringOrNull(attempt.runnerJobId);
+  const sessionId = stringOrNull(attempt.sessionId) ?? stringOrNull(record(task.sessionRef).sessionId);
+  return {
+    kind: ref.kind,
+    name: ref.name,
+    summary: {
+      id: taskId,
+      state: task.state ?? null,
+      queue: task.queue ?? null,
+      lane: task.lane ?? null,
+      backendProfile: task.backendProfile ?? null,
+      providerId: task.providerId ?? null,
+      title: task.title ?? null,
+      unread: task.unread === true,
+      failureKind: task.failureKind ?? task.degradedReason ?? null,
+    },
+    latestAttempt: {
+      attemptId: attempt.attemptId ?? null,
+      state: attempt.state ?? null,
+      runId,
+      commandId,
+      runnerJobId,
+      sessionId,
+    },
+    next: {
+      events: runId === null ? null : `bun scripts/cli.ts agentrun events run/${runId} --after-seq 0 --limit 100`,
+      logs: sessionId === null ? null : `bun scripts/cli.ts agentrun logs session/${sessionId} --tail 100`,
+      result: runId === null || commandId === null ? null : `bun scripts/cli.ts agentrun result run/${runId} --command ${commandId}`,
+      ack: `bun scripts/cli.ts agentrun ack task/${taskId}`,
+      full: `bun scripts/cli.ts agentrun describe task/${taskId} --full -o json`,
+    },
+    valuesPrinted: false,
+  };
+}
+
 function unwrapTaskDetail(data: Record<string, unknown>): Record<string, unknown> {
   const task = record(data.task);
   if (Object.keys(task).length > 0) return task;
@@ -1682,8 +1737,8 @@ function nextPagedResourceCommand(command: string, nextAfterSeq: string, limit:
   return `${parts} --after-seq ${nextAfterSeq} --limit ${limit}`;
 }
 
-function agentRunExplain(kindRaw: string): string {
-  if (kindRaw === "session-policy" || kindRaw === "provider-policy") return renderAgentRunSessionPolicyExplanation();
+function agentRunExplain(kindRaw: string, args: string[] = [], options: AgentRunRestTargetOptions = { node: null, lane: null }): string {
+  if (kindRaw === "session-policy" || kindRaw === "provider-policy") return renderAgentRunSessionPolicyExplanation(args, options);
   const kind = parseResourceKind(kindRaw);
   if (kind === "task") {
     return [
@@ -1798,7 +1853,7 @@ interface LaneConfirmOptions extends ConfirmOptions {
   lane: string | null;
 }
 
-interface RefreshOptions extends ConfirmOptions {
+interface RefreshOptions extends ConfirmOptions, DisclosureOptions {
   node: string | null;
   lane: string | null;
 }
@@ -1815,7 +1870,7 @@ interface GitMirrorOptions extends ConfirmOptions {
   wait: boolean;
 }
 
-interface CleanupRunnersOptions extends ConfirmOptions {
+interface CleanupRunnersOptions extends ConfirmOptions, DisclosureOptions {
   node: string | null;
   lane: string | null;
   timeoutSeconds: number;
@@ -2032,9 +2087,20 @@ function parseRefreshOptions(args: string[]): RefreshOptions {
   const base = parseConfirmOptions(args);
   let node: string | null = null;
   let lane: string | null = null;
+  let full = false;
+  let raw = false;
   for (let index = 0; index < args.length; index += 1) {
     const arg = args[index];
     if (arg === "--confirm" || arg === "--dry-run") continue;
+    if (arg === "--full") {
+      full = true;
+      continue;
+    }
+    if (arg === "--raw") {
+      raw = true;
+      full = true;
+      continue;
+    }
     if (arg === "--node") {
       const value = args[index + 1];
       if (value === undefined || value.startsWith("--")) throw new Error("--node requires a value");
@@ -2051,7 +2117,7 @@ function parseRefreshOptions(args: string[]): RefreshOptions {
     }
     throw new Error(`unsupported refresh option: ${arg}`);
   }
-  return { ...base, node, lane };
+  return { ...base, node, lane, full, raw };
 }
 
 function parseConfirmOptions(args: string[]): ConfirmOptions {
@@ -2078,14 +2144,17 @@ function parseGitMirrorOptions(args: string[]): GitMirrorOptions {
 }
 
 function parseCleanupRunnersOptions(args: string[]): CleanupRunnersOptions {
-  validateOptions(args, new Set(["--confirm", "--dry-run", "--force-active"]), new Set(["--timeout-seconds", "--node", "--lane"]));
+  validateOptions(args, new Set(["--confirm", "--dry-run", "--force-active", "--full", "--raw"]), new Set(["--timeout-seconds", "--node", "--lane"]));
   const base = parseConfirmOptions(args);
+  const raw = args.includes("--raw");
   return {
     ...base,
     node: optionValue(args, "--node") ?? null,
     lane: optionValue(args, "--lane") ?? null,
     timeoutSeconds: positiveIntegerOption(args, "--timeout-seconds", 180, 600),
     forceActive: args.includes("--force-active"),
+    full: raw || args.includes("--full"),
+    raw,
   };
 }
 
@@ -2598,6 +2667,91 @@ function renderAgentRunControlPlaneStatusSummary(result: Record<string, unknown>
   return renderedCliResult(result.ok !== false, "agentrun control-plane status", `${lines.join("\n")}\n`);
 }
 
+function renderAgentRunControlPlanePlanSummary(result: Record<string, unknown>): RenderedCliResult {
+  const target = record(result.target);
+  const node = record(target.node);
+  const runtime = record(target.runtime);
+  const source = record(target.source);
+  const checks = Array.isArray(result.plannedChecks) ? result.plannedChecks.map(String) : [];
+  const lines = [
+    "AGENTRUN CONTROL-PLANE PLAN",
+    renderTable(
+      ["NODE", "LANE", "NAMESPACE", "SOURCE", "CHECKS", "MUTATION"],
+      [[
+        displayValue(node.id ?? target.node ?? "-"),
+        displayValue(target.lane ?? "-"),
+        displayValue(runtime.namespace ?? "-"),
+        displayValue(source.branch ?? "-"),
+        String(checks.length),
+        "false",
+      ]],
+    ),
+    "",
+    "CHECKS",
+    ...(checks.length === 0 ? ["-"] : checks.slice(0, 12).map((check) => `- ${check}`)),
+    "",
+    "NEXT",
+    ...renderNextObjectLines(record(result.next)),
+    "",
+    "DETAIL",
+    "  use --full for rendered target details; valuesPrinted=false",
+  ];
+  return renderedCliResult(result.ok !== false, "agentrun control-plane plan", `${lines.join("\n")}\n`);
+}
+
+function renderAgentRunControlPlaneActionSummary(result: Record<string, unknown>, title: string): RenderedCliResult {
+  const target = record(result.target);
+  const node = record(target.node);
+  const runtime = record(target.runtime);
+  const summaryRows = [
+    ["ok", String(result.ok !== false)],
+    ["mode", displayValue(result.mode ?? "-")],
+    ["dryRun", displayValue(result.dryRun ?? "-")],
+    ["mutation", displayValue(result.mutation ?? "-")],
+    ["namespace", displayValue(result.namespace ?? runtime.namespace ?? "-")],
+  ];
+  const countKeys = [
+    "runnerJobCount",
+    "inactiveCandidateCount",
+    "selectedRunnerJobCount",
+    "protectedActiveRunnerCount",
+    "remainingAfterSelection",
+    "deletedRunnerJobCount",
+    "nonTerminalPodCount",
+    "forceActive",
+  ];
+  const countRows = countKeys
+    .filter((key) => result[key] !== undefined)
+    .map((key) => [key, displayValue(result[key])]);
+  const lines = [
+    title,
+    renderTable(
+      ["NODE", "LANE", "NAMESPACE", "MODE", "OK"],
+      [[
+        displayValue(node.id ?? target.node ?? "-"),
+        displayValue(target.lane ?? "-"),
+        displayValue(result.namespace ?? runtime.namespace ?? "-"),
+        displayValue(result.mode ?? "-"),
+        String(result.ok !== false),
+      ]],
+    ),
+    "",
+    renderTable(["FIELD", "VALUE"], summaryRows),
+  ];
+  if (countRows.length > 0) lines.push("", renderTable(["COUNT", "VALUE"], countRows));
+  const nextLines = renderNextObjectLines(record(result.next ?? result.followUp));
+  if (nextLines.length > 0) lines.push("", "NEXT", ...nextLines);
+  lines.push("", "DETAIL", "  use --full for capture/probe details; valuesPrinted=false");
+  return renderedCliResult(result.ok !== false, String(result.command ?? "agentrun control-plane"), `${lines.join("\n")}\n`);
+}
+
+function renderNextObjectLines(next: Record<string, unknown>): string[] {
+  return Object.values(next)
+    .filter((value): value is string => typeof value === "string" && value.length > 0)
+    .slice(0, 5)
+    .map((line) => `  ${line}`);
+}
+
 function yesNo(value: unknown): string {
   if (value === true) return "yes";
   if (value === false) return "no";
@@ -6068,9 +6222,14 @@ async function submitQueueTaskRest(args: string[]): Promise<Record<string, unkno
   const aipod = agentRunOption(args, "aipod") ?? agentRunOption(args, "aipod-spec");
   if (aipod) {
     const rendered = await agentRunRestRequest("agentrun aipod-specs render", "POST", `/api/v1/aipod-specs/${encodeURIComponent(aipod)}/render`, await aipodRenderInputFromArgs(args, 2));
-    const body = record(record(innerData(rendered)).queueTask);
+    const body = normalizeAipodRenderedQueueTask(record(record(innerData(rendered)).queueTask), args, aipod);
     if (Object.keys(body).length === 0) throw new AgentRunRestError("schema-mismatch", `aipod-spec ${aipod} render did not return queueTask`);
-    if (agentRunHasFlag(args, "dry-run")) return agentRunDryRunPlan("queue-submit", "/api/v1/queue/tasks", body, queueSubmitConfirmCommand(args, aipod), "POST", { jsonInput: { source: "aipod-spec", aipod, valuesPrinted: false } });
+    if (agentRunHasFlag(args, "dry-run")) {
+      return agentRunDryRunPlan("queue-submit", "/api/v1/queue/tasks", body, queueSubmitConfirmCommand(args, aipod), "POST", {
+        jsonInput: { source: "aipod-spec", aipod, valuesPrinted: false },
+        aipodBinding: agentRunAipodBindingDisclosure(body, aipod),
+      });
+    }
     return await agentRunRestRequest("agentrun queue submit", "POST", "/api/v1/queue/tasks", body);
   }
   const body = await requiredJsonBody(args, "queue submit");
@@ -6093,7 +6252,15 @@ async function sessionSendRest(sessionId: string, args: string[]): Promise<Recor
   const prompt = optionalPromptFromArgs(args, 2);
   const sendBody = await sessionSendBodyFromRunInput(sessionId, args, input, sessionSendPayloadFromInput(input, prompt));
   const response = await agentRunRestRequest("agentrun sessions send", "POST", `/api/v1/sessions/${encodeURIComponent(sessionId)}/send`, sendBody);
-  return { ok: response.ok !== false, command: "agentrun sessions send", data: innerData(response), bridge: response.bridge };
+  return {
+    ok: response.ok !== false,
+    command: "agentrun sessions send",
+    data: {
+      ...record(innerData(response)),
+      sessionPolicy: agentRunSessionRunPolicyDisclosure(record(sendBody.run)),
+    },
+    bridge: response.bridge,
+  };
 }
 
 function sessionSendPayloadFromInput(input: Record<string, unknown>, prompt: string | null): Record<string, unknown> {
@@ -6134,6 +6301,7 @@ function isExplicitSessionSendBody(input: Record<string, unknown>): boolean {
 async function sessionRunBodyFromArgs(sessionId: string, args: string[], input: Record<string, unknown>): Promise<Record<string, unknown>> {
   const existing = await fetchAgentRunSessionOrNull(sessionId, args);
   const sessionPolicy = readAgentRunClientConfig().client.sessionPolicy;
+  const policyTarget = resolveAgentRunSessionPolicyTarget();
   const profile = agentRunOption(args, "profile")
     ?? agentRunOption(args, "backend-profile")
     ?? stringOrNull(input.backendProfile)
@@ -6142,12 +6310,14 @@ async function sessionRunBodyFromArgs(sessionId: string, args: string[], input:
   const body: Record<string, unknown> = { ...input };
   body.tenantId = agentRunOption(args, "tenant-id") ?? stringOrNull(body.tenantId) ?? stringOrNull(existing?.tenantId) ?? sessionPolicy.tenantId;
   body.projectId = agentRunOption(args, "project-id") ?? stringOrNull(body.projectId) ?? stringOrNull(existing?.projectId) ?? sessionPolicy.projectId;
-  body.providerId = agentRunOption(args, "provider-id") ?? stringOrNull(body.providerId) ?? stringOrNull(existing?.providerId) ?? sessionPolicy.providerId;
+  body.providerId = agentRunOption(args, "provider-id") ?? stringOrNull(body.providerId) ?? stringOrNull(existing?.providerId) ?? defaultAgentRunProviderId(sessionPolicy, policyTarget);
   body.backendProfile = profile;
   body.workspaceRef = jsonObjectOption(args, "workspace-json") ?? record(body.workspaceRef);
   if (Object.keys(record(body.workspaceRef)).length === 0) body.workspaceRef = cloneJsonRecord(sessionPolicy.workspaceRef);
-  body.executionPolicy = jsonObjectOption(args, "execution-policy-json") ?? record(body.executionPolicy);
-  if (Object.keys(record(body.executionPolicy)).length === 0) body.executionPolicy = defaultAgentRunExecutionPolicy(profile, sessionPolicy);
+  const executionPolicy = jsonObjectOption(args, "execution-policy-json") ?? record(body.executionPolicy);
+  body.executionPolicy = defaultAgentRunExecutionPolicy(profile, sessionPolicy, policyTarget, {
+    basePolicy: Object.keys(executionPolicy).length === 0 ? sessionPolicy.executionPolicy : executionPolicy,
+  });
   const inheritedSessionRef = existing === null ? {} : {
     conversationId: existing.conversationId,
     threadId: existing.threadId,
@@ -6161,26 +6331,123 @@ async function sessionRunBodyFromArgs(sessionId: string, args: string[], input:
   return body;
 }
 
+function normalizeAipodRenderedQueueTask(task: Record<string, unknown>, args: string[], aipod: string): Record<string, unknown> {
+  if (Object.keys(task).length === 0) return task;
+  const sessionPolicy = readAgentRunClientConfig().client.sessionPolicy;
+  const policyTarget = resolveAgentRunSessionPolicyTarget();
+  const explicitProfile = agentRunOption(args, "profile") ?? agentRunOption(args, "backend-profile");
+  const renderedProfile = stringOrNull(task.backendProfile);
+  const fallbackProfile = sessionPolicy.backendProfile;
+  let backendProfile = explicitProfile ?? renderedProfile ?? fallbackProfile;
+  if (agentRunProviderCredentialRefs(policyTarget.spec, backendProfile).length === 0) {
+    if (explicitProfile !== null) {
+      throw new AgentRunRestError("validation-failed", `${policyTarget.configPath} has no providerCredential Secret binding for explicit --backend-profile ${explicitProfile} on ${policyTarget.spec.nodeId}/${policyTarget.spec.lane}`);
+    }
+    backendProfile = fallbackProfile;
+  }
+  if (agentRunProviderCredentialRefs(policyTarget.spec, backendProfile).length === 0) {
+    throw new AgentRunRestError("validation-failed", `${policyTarget.configPath} has no providerCredential Secret binding for AipodSpec ${aipod} backendProfile=${backendProfile} on ${policyTarget.spec.nodeId}/${policyTarget.spec.lane}`);
+  }
+  const basePolicy = record(task.executionPolicy);
+  const normalized: Record<string, unknown> = { ...task };
+  const explicitProviderId = agentRunOption(args, "provider-id");
+  normalized.tenantId = stringOrNull(normalized.tenantId) ?? sessionPolicy.tenantId;
+  normalized.projectId = stringOrNull(normalized.projectId) ?? sessionPolicy.projectId;
+  normalized.providerId = explicitProviderId ?? defaultAgentRunProviderId(sessionPolicy, policyTarget);
+  normalized.backendProfile = backendProfile;
+  if (!isRecord(normalized.workspaceRef)) normalized.workspaceRef = cloneJsonRecord(sessionPolicy.workspaceRef);
+  normalized.executionPolicy = defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy, policyTarget, {
+    includeToolCredentials: true,
+    basePolicy: Object.keys(basePolicy).length === 0 ? sessionPolicy.executionPolicy : basePolicy,
+  });
+  return normalized;
+}
+
+function agentRunAipodBindingDisclosure(task: Record<string, unknown>, aipod: string): Record<string, unknown> {
+  const policyTarget = resolveAgentRunSessionPolicyTarget();
+  const executionPolicy = record(task.executionPolicy);
+  const secretScope = record(executionPolicy.secretScope);
+  const providerCredentials = arrayRecords(secretScope.providerCredentials).map((credential) => ({
+    profile: credential.profile ?? null,
+    secretRef: {
+      name: record(credential.secretRef).name ?? null,
+      keys: Array.isArray(record(credential.secretRef).keys) ? record(credential.secretRef).keys : [],
+    },
+    valuesPrinted: false,
+  }));
+  const toolCredentials = arrayRecords(secretScope.toolCredentials).map((credential) => ({
+    tool: credential.tool ?? null,
+    secretRef: {
+      name: record(credential.secretRef).name ?? null,
+      key: record(credential.secretRef).key ?? null,
+    },
+    valuesPrinted: false,
+  }));
+  return {
+    aipod,
+    node: policyTarget.spec.nodeId,
+    lane: policyTarget.spec.lane,
+    namespace: policyTarget.spec.runtime.namespace,
+    policySource: policyTarget.source,
+    providerId: task.providerId ?? null,
+    backendProfile: task.backendProfile ?? null,
+    workspaceRef: task.workspaceRef ?? null,
+    executionPolicy: pickCompact(executionPolicy, ["sandbox", "approval", "timeoutMs", "network"]),
+    providerCredentials,
+    toolCredentials,
+    valuesPrinted: false,
+  };
+}
+
+function agentRunSessionRunPolicyDisclosure(runBody: Record<string, unknown>): Record<string, unknown> {
+  const policyTarget = resolveAgentRunSessionPolicyTarget();
+  const executionPolicy = record(runBody.executionPolicy);
+  const secretScope = record(executionPolicy.secretScope);
+  const providerCredentials = arrayRecords(secretScope.providerCredentials).map((credential) => ({
+    profile: credential.profile ?? null,
+    secretRef: {
+      name: record(credential.secretRef).name ?? null,
+      keys: Array.isArray(record(credential.secretRef).keys) ? record(credential.secretRef).keys : [],
+    },
+    valuesPrinted: false,
+  }));
+  return {
+    node: policyTarget.spec.nodeId,
+    lane: policyTarget.spec.lane,
+    namespace: policyTarget.spec.runtime.namespace,
+    policySource: policyTarget.source,
+    providerId: runBody.providerId ?? null,
+    backendProfile: runBody.backendProfile ?? null,
+    workspaceRef: runBody.workspaceRef ?? null,
+    executionPolicy: pickCompact(executionPolicy, ["sandbox", "approval", "timeoutMs", "network"]),
+    providerCredentials,
+    valuesPrinted: false,
+  };
+}
+
 async function sessionSendWithAipodRest(sessionId: string, aipod: string, args: string[]): Promise<Record<string, unknown>> {
   const rendered = await agentRunRestRequest("agentrun aipod-specs render", "POST", `/api/v1/aipod-specs/${encodeURIComponent(aipod)}/render`, await aipodRenderInputFromArgs(args, 2, { sessionId }));
   const renderedData = record(innerData(rendered));
-  const task = record(renderedData.queueTask);
+  const task = normalizeAipodRenderedQueueTask(record(renderedData.queueTask), args, aipod);
   if (Object.keys(task).length === 0) throw new AgentRunRestError("schema-mismatch", `aipod-spec ${aipod} render did not return queueTask`);
   const sessionRef = record(task.sessionRef);
   const metadata = record(sessionRef.metadata);
   const title = agentRunOption(args, "title") ?? stringOrNull(task.title);
   if (title) metadata.title = title;
   const sessionPolicy = readAgentRunClientConfig().client.sessionPolicy;
+  const policyTarget = resolveAgentRunSessionPolicyTarget();
   const backendProfile = stringOrNull(task.backendProfile) ?? sessionPolicy.backendProfile;
   const executionPolicy = record(task.executionPolicy);
   const runBody: Record<string, unknown> = {
     tenantId: stringOrNull(task.tenantId) ?? sessionPolicy.tenantId,
     projectId: stringOrNull(task.projectId) ?? sessionPolicy.projectId,
-    providerId: stringOrNull(task.providerId) ?? sessionPolicy.providerId,
+    providerId: stringOrNull(task.providerId) ?? defaultAgentRunProviderId(sessionPolicy, policyTarget),
     backendProfile,
     workspaceRef: task.workspaceRef ?? cloneJsonRecord(sessionPolicy.workspaceRef),
     sessionRef: { ...sessionRef, sessionId, metadata },
-    executionPolicy: Object.keys(executionPolicy).length === 0 ? defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy) : executionPolicy,
+    executionPolicy: Object.keys(executionPolicy).length === 0
+      ? defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy, policyTarget, { includeToolCredentials: true })
+      : defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy, policyTarget, { includeToolCredentials: true, basePolicy: executionPolicy }),
     resourceBundleRef: task.resourceBundleRef,
     traceSink: { kind: "aipod-session", aipod, sessionId, valuesPrinted: false },
   };
@@ -6196,7 +6463,19 @@ async function sessionSendWithAipodRest(sessionId: string, aipod: string, args:
   if (commandIdempotencyKey) sendBody.commandIdempotencyKey = commandIdempotencyKey;
   const response = await agentRunRestRequest("agentrun sessions send", "POST", `/api/v1/sessions/${encodeURIComponent(sessionId)}/send`, sendBody);
   const data = record(innerData(response));
-  return { ok: response.ok !== false, command: "agentrun sessions send", data: { ...data, aipod, profile: String(task.backendProfile ?? ""), valuesPrinted: false }, bridge: response.bridge };
+  return {
+    ok: response.ok !== false,
+    command: "agentrun sessions send",
+    data: {
+      ...data,
+      aipod,
+      profile: String(task.backendProfile ?? ""),
+      sessionPolicy: agentRunSessionRunPolicyDisclosure(runBody),
+      aipodBinding: agentRunAipodBindingDisclosure(task, aipod),
+      valuesPrinted: false,
+    },
+    bridge: response.bridge,
+  };
 }
 
 async function fetchAgentRunSessionOrNull(sessionId: string, args: string[]): Promise<Record<string, unknown> | null> {
@@ -6860,9 +7139,19 @@ async function aipodRenderInputFromArgs(args: string[], trailingPromptStart: num
   const input = await optionalJsonBody(args);
   const prompt = optionalPromptFromArgs(args, trailingPromptStart);
   if (prompt !== null) input.prompt = prompt;
-  copyAgentRunOptions(args, input, ["tenant-id", "project-id", "queue", "lane", "title", "provider-id", "idempotency-key", "session-id"]);
+  copyAgentRunOptions(args, input, ["tenant-id", "project-id", "queue", "node", "lane", "title", "provider-id", "idempotency-key", "session-id"]);
+  const sessionPolicy = readAgentRunClientConfig().client.sessionPolicy;
+  const policyTarget = resolveAgentRunSessionPolicyTarget({ node: stringOrNull(input.node), lane: stringOrNull(input.lane) });
+  if (input.node === undefined) input.node = policyTarget.spec.nodeId;
+  if (input.lane === undefined) input.lane = policyTarget.spec.lane;
+  if (input.providerId === undefined) input.providerId = defaultAgentRunProviderId(sessionPolicy, policyTarget);
   const profile = agentRunOption(args, "profile") ?? agentRunOption(args, "backend-profile");
-  if (profile) input.backendProfile = profile;
+  const backendProfile = profile ?? stringOrNull(input.backendProfile) ?? sessionPolicy.backendProfile;
+  input.backendProfile = backendProfile;
+  if (!isRecord(input.workspaceRef)) input.workspaceRef = cloneJsonRecord(sessionPolicy.workspaceRef);
+  if (!isRecord(input.executionPolicy)) {
+    input.executionPolicy = defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy, policyTarget, { includeToolCredentials: true });
+  }
   const priority = agentRunOption(args, "priority");
   if (priority) input.priority = Number(priority);
   const workspaceRef = jsonObjectOption(args, "workspace-json");
@@ -6924,7 +7213,18 @@ function jsonObjectOption(args: string[], flagName: string): Record<string, unkn
 }
 
 function queueSubmitConfirmCommand(args: string[], aipod?: string): string {
-  const dryRunless = args.filter((arg) => arg !== "--dry-run").join(" ");
+  const parts: string[] = [];
+  for (let index = 0; index < args.length; index += 1) {
+    const arg = args[index] ?? "";
+    if (arg === "submit" || arg === "--dry-run") continue;
+    if (arg === "--aipod" || arg === "--aipod-spec") {
+      index += 1;
+      continue;
+    }
+    if (arg.startsWith("--aipod=") || arg.startsWith("--aipod-spec=")) continue;
+    parts.push(arg);
+  }
+  const dryRunless = parts.join(" ");
   return `bun scripts/cli.ts agentrun queue submit${aipod ? ` --aipod ${aipod}` : ""}${dryRunless.length > 0 ? ` ${dryRunless}` : ""}`.trim();
 }
 
@@ -6935,15 +7235,51 @@ function jsonInputDisclosureFromArgs(args: string[]): Record<string, unknown> {
   };
 }
 
-function defaultAgentRunExecutionPolicy(profile: string, sessionPolicy: AgentRunSessionPolicyConfig = readAgentRunClientConfig().client.sessionPolicy): Record<string, unknown> {
-  const { spec } = resolveAgentRunLaneTarget({});
+function resolveAgentRunSessionPolicyTarget(options: AgentRunRestTargetOptions = { node: null, lane: null }): AgentRunSessionPolicyTarget {
+  if (activeAgentRunRestTarget !== null) {
+    return {
+      configPath: activeAgentRunRestTarget.configPath,
+      spec: activeAgentRunRestTarget.spec,
+      source: "selected-lane",
+      transport: "lane-k8s-service-proxy",
+    };
+  }
+  const node = options.node ?? null;
+  const lane = options.lane ?? null;
+  const { configPath, spec } = resolveAgentRunLaneTarget({ node, lane });
+  return {
+    configPath,
+    spec,
+    source: node !== null || lane !== null ? "selected-lane" : "default-lane",
+    transport: "direct-http",
+  };
+}
+
+function defaultAgentRunProviderId(sessionPolicy: AgentRunSessionPolicyConfig, target: AgentRunSessionPolicyTarget): string {
+  return target.source === "selected-lane" ? target.spec.nodeId : sessionPolicy.providerId;
+}
+
+function defaultAgentRunExecutionPolicy(
+  profile: string,
+  sessionPolicy: AgentRunSessionPolicyConfig = readAgentRunClientConfig().client.sessionPolicy,
+  target: AgentRunSessionPolicyTarget = resolveAgentRunSessionPolicyTarget(),
+  options: { includeToolCredentials?: boolean; basePolicy?: Record<string, unknown> } = {},
+): Record<string, unknown> {
+  const basePolicy = Object.keys(record(options.basePolicy)).length > 0
+    ? cloneJsonRecord(record(options.basePolicy))
+    : cloneJsonRecord(sessionPolicy.executionPolicy);
+  return agentRunExecutionPolicyWithLaneCredentials(basePolicy, profile, target, options.includeToolCredentials === true);
+}
+
+function agentRunExecutionPolicyWithLaneCredentials(basePolicy: Record<string, unknown>, profile: string, target: AgentRunSessionPolicyTarget, includeToolCredentials: boolean): Record<string, unknown> {
+  const spec = target.spec;
   const credentials = agentRunProviderCredentialRefs(spec, profile);
   if (credentials.length === 0) {
-    throw new AgentRunRestError("validation-failed", `config/agentrun.yaml has no providerCredential Secret binding for backendProfile=${profile} on default lane ${spec.nodeId}/${spec.lane}`);
+    throw new AgentRunRestError("validation-failed", `${target.configPath} has no providerCredential Secret binding for backendProfile=${profile} on ${target.source} ${spec.nodeId}/${spec.lane}`);
   }
   const providerCredentials = credentials.map((credential) => {
     if (credential.secretRef.namespace !== spec.runtime.namespace) {
-      throw new AgentRunRestError("validation-failed", `providerCredential ${profile} Secret ${credential.secretRef.name} is in namespace ${credential.secretRef.namespace}, expected default lane namespace ${spec.runtime.namespace}`);
+      throw new AgentRunRestError("validation-failed", `providerCredential ${profile} Secret ${credential.secretRef.name} is in namespace ${credential.secretRef.namespace}, expected target lane namespace ${spec.runtime.namespace}`);
     }
     return {
       profile: credential.profile,
@@ -6953,37 +7289,69 @@ function defaultAgentRunExecutionPolicy(profile: string, sessionPolicy: AgentRun
       },
     };
   });
-  const basePolicy = cloneJsonRecord(sessionPolicy.executionPolicy);
+  const toolCredentials = includeToolCredentials ? agentRunToolCredentialRefs(spec).map((credential) => {
+    if (credential.secretRef.namespace !== spec.runtime.namespace) {
+      throw new AgentRunRestError("validation-failed", `toolCredential ${credential.tool} Secret ${credential.secretRef.name} is in namespace ${credential.secretRef.namespace}, expected target lane namespace ${spec.runtime.namespace}`);
+    }
+    return {
+      tool: credential.tool,
+      secretRef: {
+        name: credential.secretRef.name,
+        key: credential.secretRef.key,
+      },
+    };
+  }) : [];
   const secretScope = record(basePolicy.secretScope);
   return {
     ...basePolicy,
     secretScope: {
       ...secretScope,
       providerCredentials,
+      ...(toolCredentials.length === 0 ? {} : { toolCredentials }),
     },
   };
 }
 
-function renderAgentRunSessionPolicyExplanation(): string {
+function agentRunToolCredentialRefs(spec: AgentRunLaneSpec): Array<{ tool: string; sourceId: string; secretRef: { namespace: string; name: string; key: string }; valuesPrinted: false }> {
+  return spec.secrets
+    .filter((secret) => secret.providerCredentialProfile === null && secret.id.startsWith("tool-"))
+    .map((secret) => ({
+      tool: secret.id.replace(/^tool-/u, "").replace(/-token$/u, ""),
+      sourceId: secret.id,
+      secretRef: secret.targetRef,
+      valuesPrinted: false,
+    }));
+}
+
+function renderAgentRunSessionPolicyExplanation(args: string[] = [], options: AgentRunRestTargetOptions = { node: null, lane: null }): string {
   const config = readAgentRunClientConfig();
-  const { spec } = resolveAgentRunLaneTarget({});
+  const target = resolveAgentRunSessionPolicyTarget(options);
+  const spec = target.spec;
   const sessionPolicy = config.client.sessionPolicy;
-  const credentials = agentRunProviderCredentialRefs(spec, sessionPolicy.backendProfile);
-  const execution = defaultAgentRunExecutionPolicy(sessionPolicy.backendProfile, sessionPolicy);
+  const backendProfile = agentRunOption(args, "profile") ?? agentRunOption(args, "backend-profile") ?? sessionPolicy.backendProfile;
+  const providerId = agentRunOption(args, "provider-id") ?? defaultAgentRunProviderId(sessionPolicy, target);
+  const workspaceRef = jsonObjectOption(args, "workspace-json") ?? cloneJsonRecord(sessionPolicy.workspaceRef);
+  const credentials = agentRunProviderCredentialRefs(spec, backendProfile);
+  const execution = defaultAgentRunExecutionPolicy(backendProfile, sessionPolicy, target);
   const credentialSources = credentials.map((credential) => ({
     profile: credential.profile,
     secretRef: credential.secretRef,
     valuesPrinted: false,
   }));
+  const toolCredentialSources = agentRunToolCredentialRefs(spec);
   return [
     "KIND: session-policy",
     `CONFIG: ${config.sourcePath}`,
-    `DEFAULT LANE: ${spec.nodeId}/${spec.lane}`,
-    `DEFAULTS: tenantId=${sessionPolicy.tenantId} projectId=${sessionPolicy.projectId} providerId=${sessionPolicy.providerId} backendProfile=${sessionPolicy.backendProfile}`,
-    `WORKSPACE: ${JSON.stringify(sessionPolicy.workspaceRef)}`,
+    `LANE CONFIG: ${target.configPath}`,
+    `TARGET LANE: ${spec.nodeId}/${spec.lane}`,
+    `POLICY SOURCE: ${target.source}`,
+    `TRANSPORT: ${target.transport}`,
+    `DEFAULTS: tenantId=${sessionPolicy.tenantId} projectId=${sessionPolicy.projectId} providerId=${providerId} backendProfile=${backendProfile}`,
+    `WORKSPACE: ${JSON.stringify(workspaceRef)}`,
     `EXECUTION: ${JSON.stringify(execution)}`,
     `PROVIDER CREDENTIAL SOURCES: ${JSON.stringify(credentialSources)}`,
-    "VALUES: secret payloads are not printed",
+    `TOOL CREDENTIAL SOURCES: ${JSON.stringify(toolCredentialSources)}`,
+    "VALUES: secret payloads are not printed; valuesPrinted=false",
   ].join("\n");
 }
 
@@ -7141,6 +7509,13 @@ interface AgentRunRestTarget {
   spec: AgentRunLaneSpec;
 }
 
+interface AgentRunSessionPolicyTarget {
+  configPath: string;
+  spec: AgentRunLaneSpec;
+  source: "selected-lane" | "default-lane";
+  transport: "direct-http" | "lane-k8s-service-proxy";
+}
+
 let activeAgentRunRestTarget: AgentRunRestTarget | null = null;
 
 type AgentRunBridgeCaptureBackend = "local-backend-core-broker" | "remote-frontend-websocket";