NAC 平台全量测试用例 Trace 报告

{
  "kind": "fact",
  "reason": "Record mandatory source item verdict before moving on",
  "text": "访问 Landing — PASS. Evidence: Browser opened https://nac-beta.xiaobei.top/ with title “North Agent Cloud”; visible North Agent Cloud branding, Chinese hero content, and actionable “进入控制台” and “查看文档” links. Root landing page loaded settled without visible error. Need Update: —."
}

{
  "duration_ms": 0,
  "ref": 2
}

{
  "url": "https://nac-beta.xiaobei.top/register"
}

{
  "url": "https://nac-beta.xiaobei.top/signup"
}

{
  "kind": "fact",
  "reason": "Mandatory verdict and FAIL evidence",
  "text": "注册账号 — FAIL. Issue: public self-service account registration capability is not reachable. Actual: current /login exposes only email/password 登录 and 小北 authentication, with no 注册 link/form, so username + fresh email + password cannot be submitted. Attempts: entered authentication via Landing “进入控制台”; checked the full visible login page; opened /register and /signup, both redirected to root Landing, whose header/body only link to login/control console/docs. Need Update: — (capability absent rather than merely moved/renamed)."
}

{
  "duration_ms": 0,
  "ref": 83
}

{
  "kind": "fact",
  "reason": "source-aligned verdict and created identity",
  "text": "测试项：确认 Clone；状态：PASS；证据：点击当前“使用模板”后出现“项目创建成功”密钥弹窗；底层智能体列表首行同步出现同一唯一对象 `qa-cloned-20260803-0300`（ID `83fbeb30-3f6f-4d0c-a28b-3e585c081de5`），所有者 admin、描述 `QA test clone`、最近编辑/创建均为“刚刚”。Need Update：Historical/历史：点击 Clone 弹窗确认；Current/当前：点击模板模式的“使用模板”，随后显示密钥弹窗并回到智能体列表；Suggested Update/建议更新：将确认按钮名更新为“使用模板”，并要求以列表同名首行作为持久化验证。保留测试对象名供后续版本、Playground、DevBox、部署、发布验证。"
}

{
  "duration_ms": 0,
  "ref": 68
}

{
  "duration_ms": 0,
  "ref": 8
}

{
  "kind": "fact",
  "reason": "source-aligned failing verdict",
  "text": "测试项：确认 version 已生成；状态：FAIL；Issue：成功模板克隆后未生成任何 version/draft；Actual：同一对象 `qa-cloned-20260803-0300` 的 `/workspaces/onboarding` 显示环境 0、活跃 0、部署 0 且“暂无环境”，`/workspaces/artifact-editor` 显示“暂无草稿”，不存在版本条目；Attempts：创建成功后自动进入 Onboarding 检查，再切到项目编辑器检查草稿/版本消费者视图，两个视图均为 0/空。Need Update：Historical/历史：在 Onboarding 页观察 version 列表；Current/当前：Onboarding 当前按环境/部署展示，版本/草稿位于“项目编辑器”；Suggested Update/建议更新：将验证路径更新为 Onboarding 的部署计数加项目编辑器草稿/版本列表，但保留“克隆后应生成版本”的功能预期。"
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title",
  "text": "测试项：Playground 验证\n状态：FAIL\nIssue：刚发布到 production 后，新的 Playground 会话无法得到智能体回复，并出现会话内容错配。\nActual：新建 session `sess_01KZ1Y2Z5R1QX30KJNRYBW5KNW`，发送“验证 QA modified 是否生效”；主区显示该操作员消息，但连续两次返回黄色“服务器繁忙，请稍后重试”，无 agent 回复。等待 10 秒并再次点击发送仍无成功响应；同一新 session/选中条目上方还显示旧会话“北京今天天气”及旧系统回复。\nAttempts：使用发布后的 production 目标；明确新建空会话并确认 session_id 改变；首次发送、等待 10 秒后重试；检查 URL、选中会话标题及主内容。\nNeed Update：Historical/历史：进入新 version playground 后发送“验证 QA modified 是否生效”；Current/当前：Playground 目标按 production 环境选择，并通过“新建会话”获得 session_id；Suggested Update/建议更新：改为在 production 环境新建会话发送该句，并断言当前 session 只含本会话消息且收到 `qa_modified` agent 回复。"
}

{
  "duration_ms": 0,
  "ref": 11
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title",
  "text": "测试项：回 DevBox\n状态：PASS\n证据：从 production Playground 点击侧栏“Agent4Agent 开发机”，URL 从 `/workspaces/playground?...` 切回 `/workspaces/develop?...`；页面显示绿色“运行中”、A4A 4/7 团队画布及刚发送的 `agents/agent.yaml` 评审请求追踪流，开发机上下文保留。\nNeed Update：Historical/历史：切回 develop 页；Current/当前：通过侧栏“Agent4Agent 开发机”返回，仍落在 `/workspaces/develop`；Suggested Update/建议更新：补充当前侧栏控件名，并以 develop URL、绿色运行状态及团队画布恢复为断言。"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1406,
  "y": 198
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title; testing complete",
  "text": "测试项：关闭 DevBox\n状态：PASS\n证据：在同一 disposable `qa-devbox-run-20260803-0109` 的 develop 页，关闭前显示绿色“运行中”、团队画布与顶部“停止开发机”；点击该按钮后画布/运行状态消失，页面恢复 Agent4Agent 开发机启动页，中央出现“启动 Agent4Agent 开发机”按钮，证明开发机已停止。\nNeed Update：Historical/历史：点关闭按钮；Current/当前：当前明确控件名为“停止开发机”，关闭后回到“启动 Agent4Agent 开发机”空态；Suggested Update/建议更新：把按钮名改为“停止开发机”，并要求验证绿色运行态消失且启动按钮重新出现。\nCoverage reconciliation：16 items total；PASS=10（进 DevBox、打开 agent YAML、编辑文件、保存文件、触发 A4A、触发部署、填 tag 并确认、跳到新 version Playground、回 DevBox、关闭 DevBox）；FAIL=2（终端基本可用、Playground 验证）；BLOCK=4（上传 PDF、终端确认文件存在、让 agent 用 PDF→Markdown 工具、验证输出）。"
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title",
  "text": "测试项：Playground 验证\n状态：FAIL\nIssue：刚发布到 production 后，新的 Playground 会话无法得到智能体回复，并出现会话内容错配。\nActual：新建 session `sess_01KZ1Y2Z5R1QX30KJNRYBW5KNW`，发送“验证 QA modified 是否生效”；主区显示该操作员消息，但连续两次返回黄色“服务器繁忙，请稍后重试”，无 agent 回复。等待 10 秒并再次点击发送仍无成功响应；同一新 session/选中条目上方还显示旧会话“北京今天天气”及旧系统回复。\nAttempts：使用发布后的 production 目标；明确新建空会话并确认 session_id 改变；首次发送、等待 10 秒后重试；检查 URL、选中会话标题及主内容。\nNeed Update：Historical/历史：进入新 version playground 后发送“验证 QA modified 是否生效”；Current/当前：Playground 目标按 production 环境选择，并通过“新建会话”获得 session_id；Suggested Update/建议更新：改为在 production 环境新建会话发送该句，并断言当前 session 只含本会话消息且收到 `qa_modified` agent 回复。"
}

{
  "duration_ms": 0,
  "ref": 11
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title",
  "text": "测试项：回 DevBox\n状态：PASS\n证据：从 production Playground 点击侧栏“Agent4Agent 开发机”，URL 从 `/workspaces/playground?...` 切回 `/workspaces/develop?...`；页面显示绿色“运行中”、A4A 4/7 团队画布及刚发送的 `agents/agent.yaml` 评审请求追踪流，开发机上下文保留。\nNeed Update：Historical/历史：切回 develop 页；Current/当前：通过侧栏“Agent4Agent 开发机”返回，仍落在 `/workspaces/develop`；Suggested Update/建议更新：补充当前侧栏控件名，并以 develop URL、绿色运行状态及团队画布恢复为断言。"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1406,
  "y": 198
}

{
  "kind": "fact",
  "reason": "Phase A mandatory result for exact source title; testing complete",
  "text": "测试项：关闭 DevBox\n状态：PASS\n证据：在同一 disposable `qa-devbox-run-20260803-0109` 的 develop 页，关闭前显示绿色“运行中”、团队画布与顶部“停止开发机”；点击该按钮后画布/运行状态消失，页面恢复 Agent4Agent 开发机启动页，中央出现“启动 Agent4Agent 开发机”按钮，证明开发机已停止。\nNeed Update：Historical/历史：点关闭按钮；Current/当前：当前明确控件名为“停止开发机”，关闭后回到“启动 Agent4Agent 开发机”空态；Suggested Update/建议更新：把按钮名改为“停止开发机”，并要求验证绿色运行态消失且启动按钮重新出现。\nCoverage reconciliation：16 items total；PASS=10（进 DevBox、打开 agent YAML、编辑文件、保存文件、触发 A4A、触发部署、填 tag 并确认、跳到新 version Playground、回 DevBox、关闭 DevBox）；FAIL=2（终端基本可用、Playground 验证）；BLOCK=4（上传 PDF、终端确认文件存在、让 agent 用 PDF→Markdown 工具、验证输出）。"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1120,
  "y": 981
}

{
  "text": "请描述这张图片"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1591,
  "y": 1028
}

{
  "duration_ms": 0,
  "key": "Control+R"
}

{
  "kind": "fact",
  "reason": "Formal Phase A result for source item TI-6 and decisive bug evidence",
  "text": "Test Item: 图片上传 + 用户气泡图片显示(#1369)\nStatus: FAIL\nIssue: uploaded PNG is not rendered as an image in the user bubble.\nActual: Selected available screenshot.png (169.2 kB), saw upload reach 已上传 1/1 and 100%, sent with text “请描述这张图片”; user bubble showed only an “附件” section with a filename pill screenshot.png, no thumbnail/inline image. Agent replied it could not directly view/parse the uploaded image. After Control+R the same session/message persisted but remained filename-only.\nAttempts: upload via attachment picker; send after completion; inspect user bubble and agent response; reload same session and reinspect. >1MB paste probe unavailable because no >1MB image fixture exists, but direct normal-image display violation is decisive.\nNeed Update: Historical/历史: upload or paste image.png, refresh/re-enter, plus paste >1MB image. Current/当前: current attachment picker and localized message bubble were used; no material navigation change. Suggested Update/建议更新: retain case, optionally name current “附件” bubble/file-pill states as explicit expected-vs-actual checkpoints."
}

{
  "kind": "constraint",
  "reason": "Formal Phase A result for source item TI-9",
  "text": "Test Item: Visual file inline render\nStatus: BLOCK\nAttempted: Uploaded available screenshot.png and asked agent to describe it; current agent responded it could not directly view/parse the image. Native picker contains no video file, and no known disposable agent/tool configuration exposing read_visual_file is supplied.\nBlocked by: missing required short-video fixture and unavailable known read_visual_file-capable disposable agent/configuration, so the mandated image+video tool call/output cannot be judged as a complete capability test.\nNeeded: provide a short video plus a disposable deployed agent version with read_visual_file enabled (or documented current equivalent).\nNeed Update: —"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 909,
  "y": 715
}

{
  "button": "right",
  "duration_ms": 0,
  "x": 909,
  "y": 715
}

{
  "duration_ms": 0,
  "key": "Escape"
}

{
  "kind": "fact",
  "reason": "Formal Phase A result for source item TI-10 and bug evidence",
  "text": "Test Item: 文件下载\nStatus: FAIL\nIssue: no reachable download action exists for an uploaded FileTag.\nActual: Same uploaded screenshot.png persisted in the user message as a filename pill; direct click did nothing. Right-click opened only Chrome generic page context menu (no link/file download). Latest DOM/SoM exposed no FileTag/download control, and SnapshotDrawer tree showed 0 files rather than an alternate downloadable record.\nAttempts: direct click on exact tag, browser context/right-click, DOM/SoM affordance inspection, alternate Sandbox Files drawer inspection.\nNeed Update: —"
}

{
  "duration_ms": 0,
  "ref": 14
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1100,
  "y": 980
}

{
  "text": "调用一个不存在的工具"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1590,
  "y": 1028
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: Playground 触发错误\nStatus: FAIL\nEvidence/Issue: In current Chinese “试验场” for disposable qa-devbox-run-20260803-0109 production session sess_01KZ1RBA5NNMSD3M8SQSNS24K6, sent exact prompt “调用一个不存在的工具”. The prompt appeared as a user turn at 02:56:45, but agent replied “我无法调用不存在的工具” and listed valid tools read_file/write_file/run_shell_command; no tool invocation error/failing state was produced. Direct action contradicts the intended error-trigger capability/path.\nNeed Update: Historical/历史: 进 Playground 输入「调用一个不存在的工具」可触发错误; Current/当前: Playground is labeled “试验场”, and the same prompt is safely declined rather than producing an error; Suggested Update/建议更新: update navigation label to “试验场” and replace the stale trigger prompt/setup with a current deterministic error-producing tool configuration or prompt."
}

{
  "duration_ms": 0,
  "key": "Control+V"
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: 复制\nStatus: PASS\nEvidence: Clicked the copy icon beside the expanded Input section containing “QA no-model 0219: reply OK”; then focused the session search field and pressed Ctrl+V. The exact string “QA no-model 0219: reply OK” appeared, proving clipboard output equals the selected trace input. Target session remained included.\nNeed Update: Historical/历史: generic “点复制”; Current/当前: each inspector section has its own unlabeled two-page copy icon; Suggested Update/建议更新: specify the Input section copy icon and verify by pasting into a non-destructive field."
}

{
  "duration_ms": 0,
  "ref": 8
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1795,
  "y": 609
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: 保存\nStatus: FAIL\nIssue: Trace inspector has no save capability for expanded Input/detail content.\nActual: Input, Output, and Metadata sections are read-only; each toolbar exposes only Structure/JSON toggle, fullscreen, copy, and download. Clicking Input fullscreen produced no editor or save dialog and no visible state change. No “保存” control appeared in the inspector header, section header, structured view, or JSON view.\nAttempts: inspected current Input expanded view and all section toolbars; probed fullscreen icon; checked both Structure and JSON affordances and inspector header actions.\nNeed Update: Historical/历史: “点保存” after expanding input; Current/当前: no editable/save surface exists in current trace inspector; Suggested Update/建议更新: either restore/identify the intended Save control, or remove/replace this case if trace payloads are now intentionally read-only."
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: 切暗模式\nStatus: PASS\nEvidence: Opened admin user menu, where Theme offered 浅色/深色/跟随系统 and Follow System was initially selected with a visibly light page. Clicking 深色 selected it in blue and changed header, sidebar, cards, charts, and menu to dark backgrounds with readable contrasting text.\nNeed Update: Historical/历史: generic “点主题切换”; Current/当前: theme control is inside the top-right admin user menu under “主题” with three explicit choices; Suggested Update/建议更新: open user menu, choose “深色”, and verify both selected state and full-page dark styling."
}

{
  "duration_ms": 0,
  "key": "Alt+ArrowLeft"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1850,
  "y": 140
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1825,
  "y": 270
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: 导出 JSON\nStatus: FAIL\nIssue: trace export download fails and does not deliver JSON.\nActual: Current inspector offers one “下载 trace” action rather than an export-format dropdown. Clicking it started a Chrome download named with a UUID-like value (57888315-394d-43af-a330-32abd...) but browser download bubble showed red “Something went wrong” and “Try again”; no JSON file completed.\nAttempts: located trace-level download after closing user menu; contrasted with section-level download icons and invoked the intended trace-level control on selected trace fcc7d83a.\nNeed Update: Historical/历史: trace Export → select JSON; Current/当前: single “下载 trace” button with no format chooser, and its download fails; Suggested Update/建议更新: revise navigation to “下载 trace” if JSON is now the only intended format, but fix download completion/filename first and assert a successful .json artifact."
}

{
  "ref": 7,
  "text": "version:7c9f5838"
}

{
  "duration_ms": 0,
  "ref": 5
}

{
  "ref": 6,
  "text": "7c9f5838"
}

{
  "duration_ms": 0,
  "ref": 5
}

{
  "kind": "fact",
  "reason": "mandatory item verdict",
  "text": "Test Item: 关键字搜索 session\nStatus: FAIL\nIssue: session keyword search does not index version_tag.\nActual: Same target sess_01KZ1V67SCXM0326SQZGTYADW8 was included (count 1) when searching independently by user-message fragment “QA no-model 0219”, model-output-only fragment “我是 Nex，来自 Nex-AGI”, exact sessionId, and exact traceId fcc7d83ac1d39716c41918cdaa0f5bd7. However, its visible metadata tag is version:7c9f5838; searches for both “version:7c9f5838” and raw “7c9f5838” returned count 0 / 无匹配会话.\nAttempts: cleared/replaced criteria between each named dimension, explicitly submitted Search, waited for backend sync; tested prefixed and raw version values.\nNeed Update: —"
}

{
  "text": "帮我 fetch https://httpbin.org/get"
}

{
  "duration_ms": 0,
  "key": "Enter"
}

{
  "duration_ms": 0,
  "ref": 1
}

{
  "duration_ms": 0,
  "ref": 10
}

{
  "kind": "fact",
  "reason": "Record direct functional non-conformance after distinct current-UI attempts.",
  "text": "Test item: agent fetch 外网\nStatus: FAIL\nIssue: Accessible Playground agent cannot perform requested external HTTP fetch.\nActual: In new isolated session sess_01KZ1XN5271NXXD9ZS4FXFABR8 on running deployment draft-test-45f38022-340, sent exact message “帮我 fetch https://httpbin.org/get”. Agent replied “我目前没有可用的 HTTP 请求工具，无法直接帮你 fetch https://httpbin.org/get”, returning no httpbin response data.\nAttempts: (1) Used running qa-playground agent and persisted exact request/reply; (2) checked second current project nfw-qa-20260803-0003 intended for NFW/network QA, but its consumer view shows 环境 0、活跃 0、部署 0 and 暂无环境, so no alternate Playground target exists there.\nNeed Update: —"
}

{
  "duration_ms": 0,
  "ref": 7
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 953,
  "y": 978
}

{
  "text": "北京今天天气"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1590,
  "y": 1027
}

{
  "kind": "fact",
  "reason": "source-aligned Phase A verdict",
  "text": "Test Item: Playground 验证发布版本\nStatus: FAIL\nIssue: Newly published weather-agent version is not selectable/observable in Playground and production consumer runs the base template instead.\nActual: Immediately after publishing version #0 ID ba84aa90-6600-4160-bca5-5037bb081466 to production, opened same project Playground; target selector showed production only and no version/tag selector. Sent exact “北京今天天气”; response identity was “系统智能体 (base_template_agent)” and replied it cannot directly fetch real-time weather, suggesting external sites/APIs, rather than running the generated weather agent.\nAttempts: entered Playground directly from publish success, used a new session on production, sent the exact query, and checked visible target/version controls plus response identity/content.\nNeed Update: Historical/历史: switch to Playground story and explicitly select newly published version; Current/当前: Playground selects deployment target/environment (production), not an explicit version; Suggested Update/建议更新: navigate to same-project Playground, select the environment where version was deployed, and assert response trace/agent identity corresponds to published version ID or tag."
}

{
  "duration_ms": 0,
  "ref": 25
}

{
  "duration_ms": 0,
  "ref": 25
}

{
  "duration_ms": 0,
  "key": "Control+R"
}

{
  "duration_ms": 0,
  "ref": 16
}

{
  "kind": "mistake",
  "reason": "Record mandatory source item functional failure for final report",
  "text": "Test Item: EnvVarsPanel 填必填 env\nStatus: FAIL\nIssue: Environment secret masking does not work; saved runtime values remain fully readable instead of `*****` in the UI's hidden/default state.\nActual/Evidence: Added required SQL-tool env key DB_CONN_STR with test-only value `sqlite:////tmp/qa-20260803-0305.db`, saved successfully (Runtime 1, draft 10→11 files and 17.2→17.3 KB), reloaded and confirmed persistence. The eye control was in state/label “显示密钥值” (expected hidden), yet the complete value remained visible. Toggled to “隐藏密钥值” and back to “显示密钥值”; complete value remained visible in both states, including after reload—never `*****`.\nAttempts: manual add/save; toggle show→hide and hide→show; full page reload and reopen Env Variables.\nNeed Update: Historical/历史: EnvVarsPanel 红标必填项并以 `*****` 显示 secrets; Current/当前: 低代码“环境变量”不会自动生成工具缺失项，需“添加变量”手动创建 runtime.DB_CONN_STR，提供全局眼睛按钮但隐藏态未掩码; Suggested Update/建议更新: 路径改为“工具卡缺失项→环境变量→添加变量/保存”；保留 `*****` 安全断言并据此修复当前缺陷。"
}

{
  "duration_ms": 7000,
  "load_state": "load",
  "selector": "body",
  "text": "",
  "url_pattern": ""
}

{
  "duration_ms": 7000,
  "load_state": "load",
  "selector": "body",
  "text": "",
  "url_pattern": ""
}

{
  "duration_ms": 0,
  "key": "Control+R"
}

{
  "duration_ms": 0,
  "ref": 16
}

{
  "kind": "mistake",
  "reason": "Record mandatory source item functional failure for final report",
  "text": "Test Item: 自动保存确认\nStatus: FAIL\nIssue: Low-code editor does not auto-save changes after waiting and provides no “已保存 Xs 前” confirmation.\nActual/Evidence: Changed saved DB_CONN_STR from unique `...0305.db` to `...0308.db` without clicking Save. Panel changed to “有未保存修改”; after waiting beyond several seconds it remained unsaved and no relative-time saved indicator appeared. Reloading discarded `...0308.db` and restored saved `...0305.db`, proving no automatic persistence.\nAttempts: timed wait on Env Variables valid edit; inspected panel/top status; full reload downstream readback. Other tested panels (Prompt, Env) expose explicit Save and only persisted after manual Save.\nNeed Update: Historical/历史: 等几秒查看头部“已保存 Xs 前”; Current/当前: 顶部仅静态“已保存”，各低代码面板有显式“保存”，未保存修改等待后不会自动持久化; Suggested Update/建议更新: 若产品已改为手动保存，应把用例重写为“点击面板保存→确认静态已保存→刷新回读”；若自动保存仍为需求，则保留原用例并修复。"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1000,
  "y": 981
}

{
  "text": "请务必调用 execute_sql（执行 SQL 查询）工具，执行 SELECT 1 AS qa_value，并把工具返回结果原样告诉我。"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1590,
  "y": 1030
}

{
  "duration_ms": 0,
  "ref": 24
}

{
  "kind": "mistake",
  "reason": "Record final mandatory source item failure before composing report",
  "text": "Test Item: 试跑用到 catalog tool 的对话\nStatus: FAIL\nIssue: Enabled catalog execute_sql tool is invoked but cannot execute because integration injects `db_conn_str` as a tool argument that its schema rejects.\nActual/Evidence: In a fresh Playground session targeting qa-lane-0238/current v1.0.1-qa-0308, sent explicit request to call execute_sql with `SELECT 1 AS qa_value`. Conversation showed an execute_sql tool step, then returned status:error/error_type:ValueError instead of qa_value. Opened Trace: same session had 4 spans including Tool: execute_sql; error said `Invalid parameters for tool 'execute_sql': Additional properties are not allowed ('db_conn_str' was unexpected)` and showed params including db_conn_str=`sqlite:////tmp/qa-20260803-0305.db`, max_rows=10, sql=`SELECT 1 AS qa_value`, timeout=30.\nAttempts: fresh session to avoid stale evidence; explicit named-tool prompt; waited for completed answer; opened downstream Trace and inspected actual tool span/parameters.\nNeed Update: Historical/历史: 用 PDF 示例触发 catalog tool; Current/当前: catalog 中选择并启用 SQL 查询工具，在目标泳道试验场通过明确 execute_sql 请求触发，并可由“查看 Trace”核验工具 span; Suggested Update/建议更新: 将示例改为与实际已选 catalog tool 匹配的请求，并增加“Trace 中出现对应 tool span 且结果成功”的断言。"
}

{
  "kind": "fact",
  "reason": "Formal source-item result for final report",
  "text": "api_path（IND-compat::CT-01） — BLOCK. Evidence/attempts: authenticated as admin via current /login and verified admin identity; opened Console > API 接入文档 > /docs/api-reference. Docs state /v1/chat and /v1/sessions are historical aliases mapped to same handlers, retained for compatibility, and new integrations should use /agent-api/*. Direct browser GET to https://nac-beta.xiaobei.top/v1/chat returned HTTP 405 (registered POST-only route, not 404). Blocker: browser-only harness cannot send an authenticated arbitrary HTTP POST/body and no API access key/request client was supplied, so legacy-vs-new valid response equivalence cannot be directly verified. Needed: curl/Postman/API test client plus valid NAC API key/project/version/environment IDs. Need Update: Historical/历史: old /v1/chat → new /agent-api/chat raw path action; Current/当前: compatibility alias is documented under Console > API 接入文档 > API 参考, with /agent-api/* as the public integration surface; Suggested Update/建议更新: revise case to use authenticated POSTs with identical payload against both paths and compare status, body, SSE semantics, and side effects, while citing the current API reference."
}

{
  "url": "https://nac-beta.xiaobei.top/v1/sessions"
}

{
  "kind": "fact",
  "reason": "Formal source-item result for final report",
  "text": "api_path（IND-compat::CT-02） — BLOCK. Evidence/attempts: current /docs/api-reference explicitly states /v1/sessions and /v1/chat are historical aliases mapped to the same handlers, retained for compatibility; direct GET to /v1/sessions visibly returned HTTP 405, showing the route exists but disallows GET. Blocker: no arbitrary authenticated POST/SSE client and no supplied NAC API key/project-version-environment identifiers, so creation through legacy and new endpoints plus downstream session readback cannot be compared. Needed: curl/Postman/API harness with valid access key and identifiers. Need Update: Historical/历史: old /v1/sessions → new /agent-api/sessions; Current/当前: docs group compatibility aliases in /docs/api-reference and recommend /agent-api/* for new integrations; Suggested Update/建议更新: execute identical authenticated POST payloads on both paths and verify response/session identity in the current Session/Observability consumer view."
}

{
  "url": "https://nac-beta.xiaobei.top/sessions"
}

{
  "kind": "fact",
  "reason": "Formal source-item result for final report",
  "text": "api_path（IND-compat::CT-03） — FAIL. Issue: bare /sessions does not expose a sessions API compatibility route. Actual: direct navigation to https://nac-beta.xiaobei.top/sessions redirected to https://nac-beta.xiaobei.top/ and rendered the North Agent Cloud marketing homepage HTML; by contrast /v1/sessions returned 405, proving a registered API route. Attempts: (1) direct runtime request to bare /sessions; (2) inspected current /docs/api-reference, which documents /agent-api/* and /v1/* historical aliases but no bare /sessions mapping. This is functional absence, not only stale navigation. Need Update: —."
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1100,
  "y": 972
}

{
  "text": "qa-ec05-404-20260803-0300"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1808,
  "y": 1013
}

{
  "url": "https://nac-beta.xiaobei.top/agents/e26fd61a-8df0-444d-8515-9374a0e7edca/workspaces/onboarding"
}

{
  "kind": "fact",
  "reason": "Phase A result for exact source item",
  "text": "Test Item: HTTP 404（IND-error-codes::EC-05）\nStatus: FAIL\nEvidence: Created unique disposable project qa-ec05-404-20260803-0300 (ID e26fd61a-8df0-444d-8515-9374a0e7edca); verified dedicated URL /agents/e26fd61a-8df0-444d-8515-9374a0e7edca/workspaces/onboarding and list count 3; deleted only that object through Settings > 删除项目 with exact-name confirmation; verified redirect to list, count returned to 2, object absent. Then directly revisited the saved deleted URL.\nIssue: Deleted project URL does not surface HTTP 404/not-found behavior.\nActual: Navigation was silently redirected to /workspace/agents and rendered the normal two-item workspace list.\nAttempts: Same deleted object's onboarding URL was saved before deletion and directly reopened after verified deletion.\nNeed Update: —"
}

{
  "button": "left",
  "duration_ms": 0,
  "x": 1808,
  "y": 1013
}

{
  "url": "https://nac-beta.xiaobei.top/agents/e26fd61a-8df0-444d-8515-9374a0e7edca/workspaces/onboarding"
}

{
  "kind": "fact",
  "reason": "Phase A result for exact source item",
  "text": "Test Item: HTTP 404（IND-error-codes::EC-05）\nStatus: FAIL\nEvidence: Created unique disposable project qa-ec05-404-20260803-0300 (ID e26fd61a-8df0-444d-8515-9374a0e7edca); verified dedicated URL /agents/e26fd61a-8df0-444d-8515-9374a0e7edca/workspaces/onboarding and list count 3; deleted only that object through Settings > 删除项目 with exact-name confirmation; verified redirect to list, count returned to 2, object absent. Then directly revisited the saved deleted URL.\nIssue: Deleted project URL does not surface HTTP 404/not-found behavior.\nActual: Navigation was silently redirected to /workspace/agents and rendered the normal two-item workspace list.\nAttempts: Same deleted object's onboarding URL was saved before deletion and directly reopened after verified deletion.\nNeed Update: —"
}

{
  "url": "https://nac-beta.xiaobei.top/foobar"
}

{
  "kind": "fact",
  "reason": "Phase A result for exact source item",
  "text": "Test Item: HTTP 404（IND-error-codes::EC-06）\nStatus: FAIL\nIssue: Random unknown URL does not surface HTTP 404/not-found behavior.\nActual: Direct navigation to https://nac-beta.xiaobei.top/foobar rewrote the address to https://nac-beta.xiaobei.top/ and rendered the normal landing page with 进入控制台 and 查看文档.\nAttempts: Direct current authenticated-browser navigation to the exact source path /foobar; resulting URL and page content were observed.\nNeed Update: —"
}