feat(progress-review): bind observations to current task acceptance

apps/presentation/dashboard/smoke/progress-review-evidence-smoke.mjs

@@ -0,0 +1,69 @@
+// Packaged reader + production Chat HTTP, delegation workers and independent checks.
+import assert from "node:assert/strict";
+import {spawn} from "node:child_process";
+import {once} from "node:events";
+import {createInterface} from "node:readline";
+import {mkdir} from "node:fs/promises";
+import {resolve} from "node:path";
+import {resolveTestPython} from "../../../../scripts/test-python.mjs";
+import {launchBrowser, loadPlaywright, waitForHttp} from "../../../../examples/dashboard-browser-smoke-support.mjs";
+process.env.LOOPX_PERSONAL_WORKSPACE_PACKAGED = "1";
+const {repoRoot, outputDir, port, startServer} = await import("../../../../examples/personal-workspace-browser/fixture.mjs");
+const {openWorkspacePage} = await import("../../../../examples/personal-workspace-browser/scenario-context.mjs");
+const fixture = spawn(resolveTestPython({repoRoot}), ["-u", "tests/presentation/progress_review_evidence_fixture.py"], {cwd: repoRoot, stdio: ["pipe", "pipe", "pipe"]});
+const exited = once(fixture, "exit");
+const lines = createInterface({input: fixture.stdout});
+const iterator = lines[Symbol.asyncIterator]();
+let stderr = "";
+fixture.stderr.on("data", chunk => {stderr += chunk;});
+async function next() {const line = await iterator.next(); assert.equal(line.done, false, stderr); return JSON.parse(line.value);}
+async function command(value) {fixture.stdin.write(`${value}\n`); return next();}
+let server, browser, workspace;
+try {
+  const {origin, goal_id} = await next();
+  server = await startServer();
+  const url = `http://127.0.0.1:${port}/chat/?statusUrl=/status.json`;
+  await waitForHttp(url);
+  browser = await launchBrowser(loadPlaywright().chromium);
+  workspace = await openWorkspacePage({newPage: options => browser.newPage({locale: "zh-CN", ...options})}, url);
+  const {page, api} = workspace;
+  await page.locator(".personal-goal-link", {hasText: "Product Release"}).click();
+  await mkdir(outputDir, {recursive: true});
+  await page.setViewportSize({width: 390, height: 844});
+  await page.route("**/api/chat/goal-configuration?*", async route => {
+    const response = await page.request.get(`${origin}/api/chat/goal-configuration?goal_id=${encodeURIComponent(goal_id)}`);
+    return route.fulfill({response});
+  });
+  await page.getByRole("button", {name: "Goal 设置", exact: true}).click();
+  await page.getByRole("button", {name: "能力中心", exact: true}).click();
+  await page.getByRole("button", {name: /进展评估哨兵/}).click();
+  const review = page.getByRole("region", {name: "审查证据读回"});
+  await review.getByText("绑定任务的规范验收条款", {exact: true}).waitFor();
+  await review.getByText("与目标的关系: off_goal", {exact: true}).waitFor();
+  await review.getByText("新证据增量: new_evidence", {exact: true}).waitFor();
+  await review.getByText("仅观察声明文件在两个检查点之间的净变化", {exact: true}).waitFor();
+  const coverage = review.getByText("仅观察声明文件在两个检查点之间的净变化", {exact: true});
+  await coverage.scrollIntoViewIfNeeded();
+  await review.getByText("观察范围与条款版本", {exact: true}).click();
+  await review.locator("code").scrollIntoViewIfNeeded();
+  await page.screenshot({path: resolve(outputDir, "progress-review-mixed-dimensions-mobile.png")});
+  assert.equal(await page.evaluate(() => document.documentElement.scrollWidth > window.innerWidth), false);
+  await page.setViewportSize({width: 1280, height: 960});
+  await review.scrollIntoViewIfNeeded();
+  await page.screenshot({path: resolve(outputDir, "progress-review-mixed-dimensions-desktop.png")});
+  await command("review-task-edit");
+  await review.getByRole("button", {name: "重新读取审查证据", exact: true}).click();
+  await review.getByText("条款或任务关联已变化，旧判断不可作为当前依据。", {exact: true}).waitFor();
+  await review.getByText("最新审查: stale", {exact: true}).waitFor();
+  await command("review-missing");
+  await review.getByRole("button", {name: "重新读取审查证据", exact: true}).click();
+  await review.getByText("尚无审查回执。", {exact: true}).waitFor();
+  assert.equal(await review.getByText("与目标的关系: off_goal", {exact: true}).count(), 0);
+  console.log("PASS: packaged scoped shadow readback, separate dimensions, changed-task withdrawal and missing evidence via production HTTP");
+} catch (error) {
+  if (workspace) console.error({body: (await workspace.page.locator("body").innerText()).slice(-3000), errors: workspace.errors});
+  throw error;
+} finally {
+  await workspace?.close(); await browser?.close(); server?.kill("SIGTERM");
+  fixture.stdin.end(); const [code] = await exited; lines.close(); assert.equal(code, 0, stderr);
+}

apps/presentation/dashboard/src/data/chat.ts

@@ -1650,6 +1650,20 @@ export const capabilityConfigurationEditorSchema = z.object({
   read_only_reason: z.string().optional(),
 });
 
+export const progressReviewObservationSchema = z.object({
+  mode: z.string(), read_state: z.enum(["observed", "missing", "unavailable"]), read_error: z.string().optional(),
+  receipt_count: z.number(), rejected_receipts: z.number(), stale_receipts: z.number(),
+  contract_revision: z.string().nullable(), authority: z.literal("none"),
+  latest: z.object({status: z.string(), event_id: z.string(), evidence_id: z.string(),
+    judgments: z.object({choice: z.object({relation: z.string().nullable(), increment: z.string().nullable()}).nullable(),
+      noul: z.object({serves_acceptance: z.number().nullable(), evidence_increment: z.number().nullable(), behavior_change: z.number().nullable()}).nullable()}),
+    criterion_current: z.boolean().optional(),
+    evidence_scope: z.object({coverage: z.literal("declared_file_net_change"), files: z.array(z.string()),
+      criterion_binding: z.object({origin: z.enum(["operator_study", "goal_acceptance"]), criteria_sha256: z.string(),
+        todo_id: z.string().optional(), contract_digest: z.string().optional(), contract_revision: z.number().optional(),
+        criterion_ids: z.array(z.string()).optional()})}).optional(),
+  }).nullable(),
+});
 export const capabilityConfigurationCatalogSchema = z.object({
   schema_version: z.literal("capability_configuration_catalog_v0"),
   capabilities: z.array(z.object({
@@ -1675,6 +1689,7 @@ export const capabilityConfigurationCatalogSchema = z.object({
       effective_revision: z.string(),
     }).optional(),
     documentation: z.record(z.string(), z.unknown()).optional(),
+    observation: progressReviewObservationSchema.optional(),
     context_contribution: z.object({
       supported_phases: z.array(z.enum(["before_plan", "before_delegate", "after_delegate_result"])),
       target: z.literal("coordinator"),

apps/presentation/dashboard/src/features/personal-workspace/goal-capability-settings.tsx

@@ -1,4 +1,4 @@
-import { useEffect, useMemo, useState } from "react";
+import { useEffect, useMemo, useRef, useState } from "react";
 import { AlertTriangle, Code2, LoaderCircle, RefreshCw } from "lucide-react";
 
 import {
@@ -273,6 +273,33 @@ function CapabilityCatalog({ callbacks, catalog, goalId, notification, onApplied
         {localizedSelected.capability_id === "content_ops" ? <ContentReferenceWorkbench key={goalId} /> : <CapabilityEditorStatus available={editorAvailable} t={t}
           description={readOnlyReason} />}
 
+        {selected.capability_id === "progress_review" && selected.observation ? <section className="personal-capability-observation" aria-label={locale === "zh-CN" ? "审查证据读回" : "Review evidence readback"}>
+          <h3>{locale === "zh-CN" ? "审查证据读回" : "Review evidence readback"}</h3>
+          <button type="button" onClick={onApplied}>{locale === "zh-CN" ? "重新读取审查证据" : "Recheck review evidence"}</button>
+          <p>{selected.observation.read_state === "unavailable" ? (locale === "zh-CN" ? "审查存储不可读，当前判断未知。" : "Review storage unavailable; current judgment unknown.")
+            : selected.observation.read_state === "missing" ? (locale === "zh-CN" ? "尚无审查回执。" : "No review receipts yet.")
+            : `${selected.observation.receipt_count} ${locale === "zh-CN" ? "条已读取回执" : "loaded receipts"}`}</p>
+          <p>{locale === "zh-CN" ? "过期 / 拒收" : "Stale / rejected"}: {selected.observation.stale_receipts} / {selected.observation.rejected_receipts}</p>
+          {selected.observation.latest ? <>
+            <p>{locale === "zh-CN" ? "最新审查" : "Latest review"}: {selected.observation.latest.status}</p>
+            <p>{locale === "zh-CN" ? "与目标的关系" : "Relation to goal"}: {selected.observation.latest.judgments.choice?.relation ?? "unknown"}</p>
+            <p>{locale === "zh-CN" ? "新证据增量" : "Evidence increment"}: {selected.observation.latest.judgments.choice?.increment ?? "unknown"}</p>
+            {selected.observation.latest.evidence_scope ? <>
+              <p>{selected.observation.latest.evidence_scope.criterion_binding.origin === "goal_acceptance"
+                ? (locale === "zh-CN" ? "绑定任务的规范验收条款" : "Bound to canonical task acceptance criteria")
+                : (locale === "zh-CN" ? "依据操作者研究条款；不证明规范验收关联" : "Operator study basis; canonical acceptance binding is unverified")}</p>
+              {selected.observation.latest.criterion_current === false ? <p role="alert">{locale === "zh-CN" ? "条款或任务关联已变化，旧判断不可作为当前依据。" : "Criteria or task binding changed; the old judgment is unavailable for current use."}</p> : null}
+              <p>{locale === "zh-CN" ? "仅观察声明文件在两个检查点之间的净变化" : "Observed only the net change of declared files between two checkpoints"}</p>
+              <details><summary>{locale === "zh-CN" ? "观察范围与条款版本" : "Observed scope and criterion version"}</summary>
+                <code>{selected.observation.latest.evidence_scope.criterion_binding.criteria_sha256}</code>
+                <p>{selected.observation.latest.evidence_scope.criterion_binding.criterion_ids?.join(", ")}</p>
+                <p>{selected.observation.latest.evidence_scope.files.join(", ")}</p>
+              </details>
+            </> : <p>{locale === "zh-CN" ? "旧回执未记录所选条款及文件覆盖范围。" : "Legacy receipt did not record selected criteria or file coverage."}</p>}
+          </> : null}
+          <p>{locale === "zh-CN" ? "此观察不证明任务完成；未命中漂移条件也不等于工作符合目标。" : "This observation does not prove completion; a drift condition not met does not establish on-goal work."}</p>
+        </section> : null}
+
         {localizedSelected.capability_id === "lark_event_inbox" ? (
           <section className="personal-capability-linked-setting">
             <div>
@@ -361,18 +388,23 @@ export function GoalCapabilitySettings({ callbacks, goalId, notification, onChan
   const [inspection, setInspection] = useState<GoalConfigurationInspection | null>(null);
   const [error, setError] = useState<string | null>(null);
   const [loading, setLoading] = useState(false);
+  const readGeneration = useRef(0);
 
   function load() {
     if (!goalId) return;
+    const generation = ++readGeneration.current;
     setLoading(true);
     setError(null);
+    // Withdraw the previous observation during a recheck; preserve navigation.
+    setInspection(current => current ? {...current, capability_catalog: {...current.capability_catalog,
+      capabilities: current.capability_catalog.capabilities.map(item => item.observation ? {...item, observation: undefined} : item)}} : null);
     void fetchGoalConfiguration(goalId)
-      .then(setInspection)
-      .catch((reason: unknown) => setError(reason instanceof Error ? reason.message : t("capabilities.loadFailed")))
-      .finally(() => setLoading(false));
+      .then(value => {if (generation === readGeneration.current) setInspection(value);})
+      .catch((reason: unknown) => {if (generation === readGeneration.current) setError(reason instanceof Error ? reason.message : t("capabilities.loadFailed"));})
+      .finally(() => {if (generation === readGeneration.current) setLoading(false);});
   }
 
-  useEffect(load, [goalId]);
+  useEffect(() => {setInspection(null); load(); return () => {readGeneration.current++;};}, [goalId]);
 
   if (!goalId) {
     return <p className="personal-capability-empty">{t("capabilities.chooseGoal")}</p>;

docs/architecture/agent-failure-containment-and-recovery.md

@@ -197,8 +197,7 @@ contracts; passing them does not establish live model correction, arbitrary
 remote exactly-once effects or production recovery time. Diagnostics deployment,
 overhead and long-running non-interference need separate evidence.
 
-The RFCs specify three remaining integration gaps in these partially implemented
-paths:
+The RFCs track three integration paths with separate implementation checkpoints:
 
 1. [Current-use checks for explicit delegation ancestry](rfcs/shared-goal-alignment-and-governed-amendment-v0.md#38-invalid-evidence-and-affected-consumers):
    refuse new dependent use when its declared basis is unavailable, preserving

docs/architecture/agent-failure-containment-and-recovery.zh-CN.md

@@ -88,7 +88,7 @@ Note 能保留尝试过的路径与下一步，但 summary 仍由发送者提供
 
 这些测试设计包含失败、陈旧来源、输入改变和 effect 中断等负例。Fixture provider 与一次性本地状态验证各自合同；通过不等于真实模型纠错、任意远端 exactly-once 或生产恢复时间合格。诊断部署、开销与长时间 non-interference 需要独立证据。
 
-RFC 明确了这些部分实现路径中仍需补齐的三项整合缺口：
+RFC 将以下三项整合路径分别记录为独立 checkpoint：
 
 1. [显式委派来源链的当前使用检查](rfcs/shared-goal-alignment-and-governed-amendment-v0.zh-CN.md#38-失效证据与受影响消费者)：声明依据不可用时拒绝新的依赖使用，保留历史完成，限制遍历与验证成本。
 2. [原任务修复与复验](rfcs/composable-state-machines-recovery-verification-v0.zh-CN.md#基于源码的实现顺序)：把独立失败依据传入可操作恢复路径，按当前 criterion 复验，再对账原效果。

docs/architecture/rfcs/composable-state-machines-recovery-verification-v0.md

@@ -304,15 +304,13 @@ lease/GoalRef fence, followed by M3's actual delivery callers.
 
 ### Source-grounded implementation sequence
 
-These are unimplemented composition deltas within existing M2/M3 and roadmap
-R2/R3/R4, not new milestones. Source owners were rechecked at
-`44931b6d22a50b949d43354e6ea498fb6b68d231`.
+The table tracks independent composition boundaries within M2/M3 and roadmap R2/R3/R4. Candidate checkpoints do not establish installed behavior or full M2 acceptance. The source baseline is `233cc76fd22760947d73e1501032b8b77e28148b`.
 
-| Bounded outcome | Existing entry and owner | Current gap and decisive exit |
+| Bounded outcome | Existing entry and owner | Current checkpoint and decisive exit |
 | --- | --- | --- |
 | Unavailable declared ancestry prevents new dependent use | `Delegations._read_current/start`, `delegation_results.require_dependencies/adoption_evidence`, `delegation.ts` | Alignment §3.8: source → A → B with A input invalidated must refuse current dependent use even if A output is unchanged. Qualify real read/start/adopt/settlement and packaged evidence readback; keep historical completion. |
 | Independent check failure reaches an actionable original-task recovery journey | `executor._task_validation_stage`, `ValidatedTurnReceipt`, canonical `turn_loop_controller_contract_v0.json`, `turn_journal.ts` | Preserve qualified failure scope and repair/replan detail; complete bounded repair or verifier-only retry, current validation and original-effect settlement. Host assertions cannot become trusted validation. |
-| Optional semantic review explains its evidence and coverage | Existing progress-review receipt/loader and canonical Goal acceptance inspection | Bind a selected criterion and evidence coverage in shadow readback; show missing/stale basis and separate judgment dimensions. Model quality and intervention qualify separately. |
+| Shadow review explains criteria and coverage | progress-review receipt/context, `progress_review_evidence.ts`, canonical acceptance inspect | Select current canonical task criteria; exact GoalRef preserves instance identity, while changed tasks or recreated instances withdraw old judgments. Independent older cores retain only the legacy manual-study shape; canonical scope never downgrades. Show separate dimensions, declared-file net change, missing evidence and unreadable storage. Default off and existing assist trigger rules retain their semantics. |
 
 For the second outcome, `_task_validation_stage` already saves the independent
 result and blocks settlement. `ValidatedTurnReceipt` omits that validation's
@@ -345,6 +343,15 @@ complete the journey. Bound repeated verification and source-chain traversal
 using measured workloads. Roll back code through its owner while retaining
 receipts, completed effects and unresolved recovery obligations.
 
+An optional `acceptance_scope` refers only to registry/runtime paths inside the
+selected workspace. Criteria come from the current owner; private commands and
+paths stay out of model questions. Hashes, versions and declared coverage are
+observations, not proof of model correctness, a durable business checkpoint or
+whole-task completion. Existing team evidence and capability settings carry App
+readback, and CLI/MCP share the owner; no Lark-specific protocol is introduced.
+Live correction, remote exactly-once, takeover, observer quality and intervention
+retain their existing RFC acceptance owners.
+
 ## 12. Open decisions
 
 The first M2 implementation must choose its smallest sufficient exploration