@tea-agent/loop-agent 0.29.2 → 0.29.3

package/docs/templates/frontend-test-dag.json

@@ -11,7 +11,7 @@
   "globalConstraints": [
     "Do not generate pytest or Playwright source code.",
     "Only use declared isolated test environments; production URLs and real credentials are blocked.",
-    "Every generated browser start command uses playwright-cli open --browser=chrome --headless followed by the concrete controller-resolved URL (from task source config.md when present, else http://localhost:5173); executable case lines never retain an angle-bracket URL/ref placeholder; subsequent commands stay in that default session and must not use unverified named-session flags.",
+    "Every generated browser start command uses playwright-cli open --browser=chrome followed by the concrete controller-resolved URL (from task source config.md when present, else http://localhost:5173); executable case lines never retain an angle-bracket URL/ref placeholder; subsequent commands stay in that default session and must not use unverified named-session flags.",
     "Case children execute serially. Persist each case result, logs and browser evidence before the next child starts.",
     "A token threshold is a post-case stop check, not a model hard token cap; unstarted cases must be recorded as blocked: token-budget-exhausted.",
     "Default pipeline acceptance is the final frontend-test-result-v1 plus testcase/frontend/reports/frontend-test-report.html",

package/docs/templates/frontend-test-dag.retrieve-context.prompt.md

@@ -10,7 +10,7 @@ The preflight controller resolves and freezes one absolute browser base URL from
 - Never use production hosts or credentials.
 - Write `environmentProbe: pending`. The environment preflight shell will replace this with `reachable`, `unreachable`, or `curl-unavailable` plus a structured `blockedReason` (for example `frontend-base-url-unreachable`).
 - Include the exact browser start prefix using the supplied controller-frozen URL:
-  `playwright-cli open --browser=chrome --headless <controller-frozen-base-url>`.
+  `playwright-cli open --browser=chrome <controller-frozen-base-url>`.
 
 Do not claim the environment is reachable until preflight completes. Preflight does not start the application.
 

package/docs/templates/frontend-test-dag.review-cases.prompt.md

@@ -2,4 +2,4 @@
 
 Read the RAG files and Markdown cases only. First line must be `VERDICT: pass` or `VERDICT: request-revision`. Report AC coverage, case independence, evidence completeness, unsafe environment/data dependencies, and manifest issues. This verdict is a deterministic safety gate: `request-revision` blocks manifest materialization and browser execution.
 
-Every case must retain the exact browser-start command prefix with the resolved absolute baseUrl from `testcase/frontend/rag/context.md` (prefer task source `config.md`, else `http://localhost:5173`): `playwright-cli open --browser=chrome --headless <resolved-base-url>`; session flags must not precede `open`, and subsequent commands must remain in its default session without `-s=` or assumed named-session binding. Verify every executable sub-scenario specifies fixture/reset, UI reset, fresh snapshot before element refs, a successful `playwright-cli find ...` semantic assertion after `open`, and an evidence write point. `snapshot`, `goto`, `screenshot`, `request`/`console`, and ordinary interactions cannot independently authorize passed. Verify each case requires both `execution.md` and `case-result.json` under its own evidence directory. The JSON result must contain matching `caseId`, `status` (`passed`, `failed`, or `blocked`) and `evidencePaths`; blocked cases must name a non-empty `blockedReason` and cannot count as passed.
+Every case must retain the exact browser-start command prefix with the resolved absolute baseUrl from `testcase/frontend/rag/context.md` (prefer task source `config.md`, else `http://localhost:5173`): `playwright-cli open --browser=chrome <resolved-base-url>`; session flags must not precede `open`, and subsequent commands must remain in its default session without `-s=` or assumed named-session binding. Verify every executable sub-scenario specifies fixture/reset, UI reset, fresh snapshot before element refs, a successful `playwright-cli find ...` semantic assertion after `open`, and an evidence write point. `snapshot`, `goto`, `screenshot`, `request`/`console`, and ordinary interactions cannot independently authorize passed. Verify each case requires both `execution.md` and `case-result.json` under its own evidence directory. The JSON result must contain matching `caseId`, `status` (`passed`, `failed`, or `blocked`) and `evidencePaths`; blocked cases must name a non-empty `blockedReason` and cannot count as passed.

package/docs/templates/init-managed-agents.md

@@ -32,10 +32,10 @@
 | --- | --- |
 | **允许** | 已发布 `loop-agent` / `agent-worker` CLI；只读 status/doctor/report/inspect/observe；准备 `source/*` 与 `task.json` 边界；human gate；shell 验证与 handoff。 |
 | **禁止** | 绕过 CLI 用宿主 Edit/Write/ApplyPatch 直接改业务实现；CLI/DAG 失败后「救火改文件」；用聊天自述代替 shell 验证。 |
-| **失败时** | `dag doctor` / `dag report` / `dag reconcile-run`（及适用 worker reconcile）；修正任务源/`task.json`/DAG 后经 CLI 重跑。 |
+| **失败时** | 先 `dag report` / `dag doctor`；优先 `dag rerun --from-node` 安全子图续跑；仅契约/源真变或 plan 不合格时同 task 重 advance / `dag rerun-task`；paused 用 approve→`dag resume`。 |
 | **实现写入** | 业务代码 **只** 经受治理 DAG writer（`implement-pi` / `repair-pi`）经 `task advance`（批准 writeSet gate 后长跑）。 |
 
-**永远不要**：`loop-agent` / `agent-worker` 失败 ⇒ 主会话直接改仓库实现。
+**永远不要**：`loop-agent` / `agent-worker` 失败 ⇒ 主会话直接改仓库实现；也勿因 provider 抖动或只读节点失败**新建无关 task-id**。
 
 ### 自然语言入口路由
 
@@ -47,6 +47,7 @@
 | 初始化对齐 / 升级后对齐 / reconcile 初始化 / loop-agent 初始化对齐 | 初始化升级闭环 | 直接运行 `init upgrade`，而非拼接低层 safe-update 命令 |
 | 初始化安全更新 / 应用初始化更新 / loop-agent 初始化安全更新 / loop agent初始化安全更新 | 初始化升级闭环 | 直接运行 `init upgrade`，由 controller 处理安全更新与验证 |
 | loop-agent 帮我完成 / 帮我实现 / 帮我修复 / 帮我开发 <需求>；使用 loop-agent 完成 <X>；按 loop-agent 流程处理 <X> | 通用需求实现 | 先提供 PRD 与 `allowedPaths` / `forbiddenPaths` / verify，再 `task advance` → 审查 writeSet gate → `task advance --approve-gate`；主会话不得直接修改业务实现 |
+| 从失败节点继续 / 续跑 / 不要重开任务 / 从 plan 或 code-review 接着跑 | DAG 中途恢复 | **不是**新 `task advance`：`dag report` → `dag rerun --from-node … --plan` → 带 `plan-hash` 执行（见失败默认恢复序） |
 
 **初始化升级闭环（写入型）**：主会话收到上述写入型表达时自动调用统一入口，并在 controller 返回的具体单文件 `allowedPaths` 内完成必要语义合并后使用 `--continue`。不得停在 `check-update`、`needs-safe-update`、`needs-model-merge` 或 `verification-pending`。默认安装/合并项目 `.opencode/plugins/`、`.pi/extensions/` 与 `.pi/settings.json`；Pi 必须 trust 项目才能加载，默认严禁写 `~/.pi/agent/settings.json`。
 
@@ -136,7 +137,9 @@ agent-worker console serve --repo . --port 8790   # 兼容入口，等价于上
 
 live run 先用 `loop-agent dag status --run-id <run-id>` 看 lifecycle 与 liveness；用 `loop-agent dag report --run-id <run-id> --markdown` 读 facts；失败/paused 用 `loop-agent dag doctor --run-id <run-id> --markdown`。生命周期对齐先只读运行 `loop-agent dag reconcile-run --run-id <run-id>`；只有 runner 已停止且 operator 明确提供 `--action supersede|abandon --reason "..."` 时才允许变更。失败 run 用 `loop-agent dag closeout-draft --run-id <run-id>` 生成 failure handoff，不要写成成功 closeout。
 
-Operator 须持续监控 live run，直到 controller 报告 run 已结束（节点/流程终态如 `FINISHED`、`FAILED` 或 `partial_failed`），或 Decision Gate **需要 approve**；不要在节点仍运行时假定完成。持续监视过程中，主会话可在合适节点（例如节点/rank 状态变化、进入 verify/closeout、出现 stall 嫌疑或需要 approve 时）向用户做简短进度汇报（当前节点、状态、是否有风险），避免长时间静默；汇报是告知，不是请求确认，不得因此停下流程。判活须组合 runner heartbeat、session events 与 `dag doctor` liveness/provider meaningful progress；Runner heartbeat 只证明 lease，alone ≠ progress，不得仅凭运行时长结束节点。exclusive writer（如 `implement-pi` / `repair-pi`）运行期间：主会话与其他 writer **不得并发修改工作区**，以免 write-guard 错误归因；只读 status/doctor/report/observe 与 approve/reject/resume CLI 仍允许。恢复：status/doctor/report → classify → reconcile/replan → CLI 重跑 → shell verify。**禁止**把主会话直接 Edit 业务代码当作恢复手段。
+**失败默认恢复序**：`dag report`/`dag doctor` → 优先 `dag rerun --from-node <node> --plan` 再带 `--plan-hash`（provider 抖动、plan/review/verify 安全下游；writer/decision/fingerprint 不合格勿硬跑）→ paused：approve→`dag resume` → 契约/源真变或 R1 不合格：同 task `task advance` / `dag rerun-task`（禁无理由新建 task-id）→ Worker-owned：`agent-worker task retry`。
+
+Operator 须监控 live run 至终态（FINISHED / FAILED / partial_failed）或 Decision Gate 需要 approve；可在节点/rank 变化、verify/closeout、stall 或需 approve 时简短汇报（告知非请求确认）。判活须组合 runner heartbeat、session events 与 `dag doctor` liveness/provider meaningful progress（heartbeat alone ≠ progress）。bounded writer 运行期间不得并发修改工作区（write guard / write-guard 会把越界 diff 错误归因到 writer）；只读 status/doctor/report 与 approve/reject/resume 仍允许。恢复：report/doctor → **优先** `dag rerun --from-node` → 必要时 rerun-task/同 task advance → shell verify。**禁止**把主会话直接 Edit 业务代码当作恢复手段。
 
 ### 运行态与验证
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tea-agent/loop-agent",
-  "version": "0.29.2",
+  "version": "0.29.3",
   "type": "module",
   "bin": {
     "loop-agent": "bin/loop-agent.js",

package/skills/playwright-cli/SKILL.md

@@ -58,7 +58,7 @@ request
 获取 snapshot，并只使用 snapshot 中可见的 ref 或已知安全 locator。
 
 ```text
-playwright-cli open --browser=chrome --headless http://localhost:5173
+playwright-cli open --browser=chrome http://localhost:5173
 playwright-cli snapshot
 playwright-cli find "Sign in"
 playwright-cli fill e1 "user@example.com"

package/skills/playwright-cli-case-generator/SKILL.md

@@ -41,7 +41,7 @@ description: 根据 FE-test RAG 知识包生成可由受限 playwright_cli runti
    `-s=<case-id>` 建立 named session。
 3. 可独立执行的命令序列：使用 RAG `context.md` 中已解析的绝对 `baseUrl`（优先来自任务源
    `config.md`；缺失时默认 `http://localhost:5173`），必须以
-   `playwright-cli open --browser=chrome --headless` 后接解析出的具体 base URL 开始；不得保留
+   `playwright-cli open --browser=chrome` 后接解析出的具体 base URL 开始；不得保留
    base-url 占位符，也不得使用 `-s=<case-id>` 或其他 named session；再按需登录、数据准备、
    `snapshot` 后优先使用元素引用和操作。每个拟通过场景必须在 `open` 后以成功的 `find` 作为
    唯一结构化 UI 语义断言，再由 controller 执行 post-execution cleanup。`snapshot`、`goto`、