@tea-agent/loop-agent 0.4.0 → 0.6.0

package/dist/workflows/dag/validate.js

@@ -428,7 +428,7 @@ function validateDecisionGateTaskConfig(task, issues) {
         });
     }
 }
-export function collectExecutorModelWarnings(spec) {
+export function collectExecutorModelWarnings(spec, expectedMatrices = DEFAULT_DAG_EXECUTOR_MODELS) {
     const warnings = [];
     const executors = Object.keys(DEFAULT_DAG_EXECUTOR_MODELS);
     for (const executor of executors) {
@@ -436,11 +436,12 @@ export function collectExecutorModelWarnings(spec) {
         if (!actualByComplexity) {
             continue;
         }
+        const expectedByComplexity = expectedMatrices[executor] ?? DEFAULT_DAG_EXECUTOR_MODELS[executor];
         for (const [complexityKey, actual] of Object.entries(actualByComplexity)) {
             if (actual === undefined) {
                 continue;
             }
-            const expected = DEFAULT_DAG_EXECUTOR_MODELS[executor][complexityKey];
+            const expected = expectedByComplexity[complexityKey];
             if (actual !== expected) {
                 warnings.push({
                     executor,

package/docs/README.md

@@ -51,6 +51,8 @@
 - `templates/exec-plan.md` — 非平凡工作的执行计划
 - `templates/progress-log.md` — 进度与交接日志
 - `templates/qa-report.md` — 验证与 QA 证据
+- `templates/worker-dogfood-setup.md` — 发布控制器下的真实 Worker sample setup 与 retry 纪律
+- `templates/worker-dogfood-evidence.md` — BE/FE/QA sample、Observe、morning report 与 coverage evidence 模板
 - `templates/production-readiness-checklist.md` — 低/中风险单仓库 DAG readiness 检查清单
 - `templates/init-evolution-review.md` — 初始化能力演化审查报告模板
 - `templates/adr.md` — 架构决策记录（ADR）

package/docs/architecture/runtime-boundaries.md

@@ -20,6 +20,10 @@ Workflow runtime (src/workflows/)
 Executors (src/executors/)
   └─ Cursor / Pi / Shell 等外部工具适配；不应依赖 commands 或 CLI formatting
 
+Worker adapter (src/worker/)
+  └─ 产品线 TaskSpec / Task Pool / Observe 本地适配；以子进程调用已发布 loop-agent CLI，
+     不得 in-process import CLI、commands 或 application
+
 Infrastructure / Store (src/infrastructure/，逐步引入)
   └─ .harness 文件系统副作用、run lifecycle、原子写入规则
 
@@ -63,6 +67,13 @@ Governance (scripts/check-*.sh, src/governance/)
 - **允许依赖**：`src/shared/**`、外部 SDK。
 - **禁止**：依赖 `src/commands/**` 或 CLI 输出格式。
 
+### Worker adapter
+
+- **位置**：`src/worker/**`，独立 `agent-worker` CLI 为 `src/worker/cli.ts`。
+- **职责**：TaskSpec 校验与物化、Task Pool batch/retry/morning report、失败路由，以及只读 Observe 事件/快照/UI；实际 DAG 执行通过 `LoopAgentClient` 启动已发布的 `loop-agent` 子进程。
+- **允许依赖**：TaskSpec、Task Pool、observability、Node filesystem/path 与明确的 shared/task contract；它不是第二套 executor 或 DAG kernel。
+- **禁止**：in-process import `src/cli/**`、`src/commands/**` 或 `src/application/**`。
+
 ### Infrastructure / Store
 
 - **位置**：`src/infrastructure/harness/**`（按计划逐步引入）；过渡期部分逻辑仍在 `src/workflows/dag/lifecycle.ts`、`src/records/**`。
@@ -89,6 +100,9 @@ Workflow runtime      ──X──>        Commands  （目标态禁止；见
 Executors             ──────────>   Shared / 外部 SDK
 Executors             ──X──>        Commands / CLI
 
+Worker adapter        ──────────>   TaskSpec / Task Pool / observability / 子进程 CLI contract
+Worker adapter        ──X──>        Commands / CLI / Application（in-process）
+
 Application           ──────────>   Workflows / Infrastructure / Task / Records
 Application           ──X──>        CLI formatting
 
@@ -103,14 +117,15 @@ Runner / Loop         ──(迁移中)──>  逐步改为仅经 Store / Appli
 | `src/commands/**` | application, workflows, infrastructure, task, records, executors, shared | — |
 | `src/workflows/**` | executors, task, records, shared, application（目标） | `src/commands/**` |
 | `src/executors/**` | shared, 外部 SDK | `src/commands/**`, `src/cli/**` |
+| `src/worker/**` | worker-local modules, task contracts, shared, Node APIs | `src/commands/**`, `src/cli/**`, `src/application/**` |
 | `src/cli/**` | commands, adapters, shared, application（目标） | — |
 | `skills/**` | （Markdown 引用，非 TS import） | 运行时 TS 模块 |
 
 ## 已知过渡耦合
 
-当前没有允许的 `src/workflows/**` 或 `src/executors/**` → `src/commands/**` TypeScript import。
+当前没有允许的 `src/workflows/**` / `src/executors/**` forbidden import，也没有允许的 `src/worker/**` → `src/{cli,commands,application}/**` TypeScript import。
 
-`scripts/check-architecture-boundaries.sh` 的 transitional allowlist 保持为空。任何新增的 `workflows/executors -> commands` import 必须导致检查 **exit 1**；如果未来确有临时例外，必须先写入 active exec plan，说明移除时间和验证门禁。
+`scripts/check-architecture-boundaries.sh` 的 transitional allowlist 保持为空。任何新增的 `workflows/executors/worker` forbidden import 必须导致检查 **exit 1**；如果未来确有临时例外，必须先写入 active exec plan，说明移除时间和验证门禁。
 
 ## Governance 钩子
 
@@ -118,7 +133,7 @@ Runner / Loop         ──(迁移中)──>  逐步改为仅经 Store / Appli
 
 | Script | 检查内容 | 失败条件 |
 |--------|----------|----------|
-| `scripts/check-architecture-boundaries.sh` | `workflows` / `executors` → `commands` import | 新的未 allowlist violation |
+| `scripts/check-architecture-boundaries.sh` | workflow/executor forbidden import，及 Worker → CLI/commands/application import | 新的未 allowlist violation |
 | `scripts/check-command-registry-drift.sh` | `command-reference.md` 中的 top-level command vs `src/cli/catalog.ts` | 文档引用未注册 command |
 | `scripts/check-skill-entry.sh` | `SKILL.md` reference 文件存在、行数阈值 | reference 缺失（fail）；行数 > 220（warn） |
 

package/docs/design/README.md

@@ -13,17 +13,29 @@
 | `六个月规划.md` | 六个月路线图与目标架构 |
 | `taskspec-to-loop-agent-mapping.md` | 将产品线 TaskSpec 适配为 `loop-agent` task 的契约 |
 | `state-and-failure-taxonomy.md` | 文档、Task Pool、DAG、Loop 共用的 canonical status 与 failure taxonomy |
+| `observe-ui.md` | 已实现的 agent-worker 本地只读可观测 UI（事件流 / SSE / stale 诊断）设计、事件契约与验收边界 |
+| `observe-ui-optimization.md` | 2026-07-10：**已实现** Observe UI 中文化、DAG-first 总览、进行中 DAG 置顶、read-model enrichment、Pi `session-events.jsonl` 落盘与过程时间线（UI-1～UI-9） |
+| `archive/2026-07-10-下一阶段任务-功能开发完成.md` | 已归档：真实样本、失败硬化与 QA 闭环的功能性开发完成记录 |
 
-## 当前实现边界
+## 命名约定（2026-07-09 起）
 
-`loop-agent` 当前拥有：仓库本地 harness、Agent DAG 生成/验证/执行、shell 验证、artifacts、reports、Loop state。它不是产品线 Task Pool、Orchestrator、docs-sync 服务、Git/CI 平台或 artifact store。
+| 名称 | 含义 | 状态 |
+|---|---|---|
+| `loop-agent` | 仓库本地 runtime / 治理控制器 CLI | 已发布 |
+| `agent-worker` | 产品线 TaskSpec 批处理适配层 CLI（`bin/agent-worker.js`） | 已发布（0.5.0） |
+| `agent-platform` | 早期规划设想名；**不是**当前 CLI。若出现在六个月目标架构中，仅表示未来产品线 monorepo / 上层服务集合，不得写成可执行命令 |
 
-设计文档使用两类 profile 名称：
+当前真实分层：
 
-- **Business task profiles**：`backend-feature`、`frontend-feature`、`qa-casegen`、`qa-testcode`、`reviewer-gate`
-- **`loop-agent dag run-task --profile` governance profiles**：`auto`、`minimal`、`standard`、`reviewed`、`supervised`
+```text
+产品线 docs / TaskSpec / TaskGraph
+  → agent-worker（本地批处理适配）
+    → loop-agent（runtime kernel）
+      → Cursor / Pi / shell / static
+```
+
+命令示例一律使用 `agent-worker`，不要再写 `agent-platform task ...`。
 
-除非 CLI 显式扩展支持，否则不要把 business task profile 直接传给 `loop-agent dag run-task --profile`。用 `taskspec-to-loop-agent-mapping.md` 将 business profile 路由到当前 governance profile。
 
 ## 维护规则
 
@@ -31,6 +43,7 @@
 
 1. 保持 TaskSpec 字段、status 名称、failure category 与命令示例与契约文档一致
 2. 未来平台组件标为目标架构，除非已存在于 `src/` 或 `scripts/`
-3. `src/task/config-types.ts` 或 DAG profile routing 变更时更新 `taskspec-to-loop-agent-mapping.md`
-4. DAG report category、Loop failure class 或 Task Pool state 变更时更新 `state-and-failure-taxonomy.md`
-5. 运行 `bash scripts/check-repo.sh`
+3. 当前 CLI 命令示例只用 `agent-worker` / `loop-agent`；`agent-platform` 仅可出现在“未来 monorepo/上层服务”语境，并显式标注未实现
+4. `src/task/config-types.ts` 或 DAG profile routing 变更时更新 `taskspec-to-loop-agent-mapping.md`
+5. DAG report category、Loop failure class 或 Task Pool state 变更时更新 `state-and-failure-taxonomy.md`
+6. 运行 `bash scripts/check-repo.sh`

package/docs/exec-plans/active/README.md

@@ -4,4 +4,9 @@
 
 源码仓库可在本 README 旁保留具体 active plan 文件。npm 包只携带本 README 作为目录契约，不复制 loop-agent 源码历史的 active plan；目标仓库自行生成 active plan。
 
-当前没有 active plan。`2026-07-08-taskspec-worker-master.md` 的 7 个 step 已全部完成并归档到 `../completed/`。
+`2026-07-08-taskspec-worker-master.md` 的 7 个 step 已全部完成并归档到 `../completed/`。
+
+当前没有 active execution plan 以外的进行中项。
+
+- active：`2026-07-10-release-0.6.0-nightly-drill.md` — 发布 0.6.0，并用已发布控制器完成第 4 周首轮夜间 batch + 次日 morning report 决策闭环。
+- 已归档：`../completed/2026-07-10-next-stage-worker-evidence.md` 及更早的 TaskSpec/OBS 计划。

package/docs/exec-plans/completed/README.md

@@ -7,6 +7,7 @@ npm 包携带本 README 作为目录契约。具体 completed plan 属于目标
 - [`2026-07-02-loop-agent-subject-restructure.md`](2026-07-02-loop-agent-subject-restructure.md) — 将原 `tools/code-agent` runtime 提升到仓库根、重命名为 `loop-agent`，移除旧 memory plugin 产品线
 - [`2026-07-04-remove-level1-fallback.md`](2026-07-04-remove-level1-fallback.md) — 移除历史顺序 Level 1 fallback，runtime、文档与 command surface 收敛到 DAG 执行
 - [`2026-07-04-runtime-boundary-remediation.md`](2026-07-04-runtime-boundary-remediation.md) — 整合 CLI/skill/runtime 边界，抽出 DAG/Loop runtime seam，集中 harness store/guard 策略
+- [`2026-07-10-next-stage-worker-evidence.md`](2026-07-10-next-stage-worker-evidence.md) — Worker retry、EnvFailure、边界治理、初始化投影与真实 BE/FE/QA dogfood evidence 闭环
 - [`2026-07-04-dag-role-skill-alignment.md`](2026-07-04-dag-role-skill-alignment.md) — 对齐 DAG/Dynamic Workflow role 与 repo-local vetted skills，新增 strict skill audit
 - [`2026-07-06-production-readiness-hardening.md`](2026-07-06-production-readiness-hardening.md) — 冻结 Production Readiness v0.1，打磨 DAG 主路径 next steps、failure routing、doctor/report/failure handoff 与 dogfood 验证
 - [`2026-07-08-taskspec-schema-validate.md`](2026-07-08-taskspec-schema-validate.md) — 新增 TaskSpec v0.1 schema、三层校验器、risk→complexity 映射和 5 个 dogfood TaskSpec 样例
@@ -17,3 +18,14 @@ npm 包携带本 README 作为目录契约。具体 completed plan 属于目标
 - [`2026-07-08-run-task-pipeline-ring2.md`](2026-07-08-run-task-pipeline-ring2.md) — 新增 Worker run-task pipeline、task-local run record、成功/失败分流与 Ring 2 shell-only DAG success/failure smoke
 - [`2026-07-08-task-pool-morning-report.md`](2026-07-08-task-pool-morning-report.md) — 新增 Worker Task Pool JSONL/state store、failure routing、串行 batch run-ready 和 morning report
 - [`2026-07-08-taskspec-worker-master.md`](2026-07-08-taskspec-worker-master.md) — 完成 TaskSpec + Worker 7 步 master 编排，覆盖 schema、graph、profile、client、materialize、run-task、Task Pool 和 morning report
+- [`2026-07-10-obs-001.md`](2026-07-10-obs-001.md) — WorkerEvent schema 与 append-only event/snapshot store
+- [`2026-07-10-obs-002.md`](2026-07-10-obs-002.md) — LoopAgentClient 流式输出、heartbeat 与实时 artifact
+- [`2026-07-10-obs-003.md`](2026-07-10-obs-003.md) — composite progress 与 Worker pipeline 事件打点
+- [`2026-07-10-obs-004.md`](2026-07-10-obs-004.md) — observability snapshot read model
+- [`2026-07-10-obs-005.md`](2026-07-10-obs-005.md) — 本地只读 observe HTTP/SSE server
+- [`2026-07-10-obs-006.md`](2026-07-10-obs-006.md) — Dashboard、Batch Detail 与 Run Detail UI
+- [`2026-07-10-obs-007.md`](2026-07-10-obs-007.md) — run-dag JSONL events 与 composite observer
+- [`2026-07-10-obs-008.md`](2026-07-10-obs-008.md) — Worker DAG events 接线与 DAG Detail UI
+- [`2026-07-10-obs-009.md`](2026-07-10-obs-009.md) — stale、quiet 与 timeout-risk 诊断
+- [`2026-07-10-obs-010.md`](2026-07-10-obs-010.md) — Failure Inbox、文档、package surface 与 smoke 验证
+- [`2026-07-10-observe-ui-review-remediation.md`](2026-07-10-observe-ui-review-remediation.md) — 修复 Observe UI 事件链路、历史 run 投影、artifact 安全边界与失败状态展示

package/docs/init-surface.manifest.json

@@ -14,6 +14,8 @@
     "docs/skills/vetted-skill-registry.md",
     "docs/templates/init-evolution-review.md",
     "docs/templates/production-readiness-checklist.md",
+    "docs/templates/worker-dogfood-setup.md",
+    "docs/templates/worker-dogfood-evidence.md",
     "docs/templates/agent-dag.schema.json",
     "examples/example-dag.json",
     "skills/loop-agent/SKILL.md",
@@ -31,6 +33,7 @@
   "initFullRequired": [
     "README.md",
     "AGENTS.md",
+    ".gitignore",
     "harness.json",
     "docs/README.md",
     "docs/development-principles.md",
@@ -39,6 +42,8 @@
     "docs/loop-agent-harness.md",
     "docs/templates/init-evolution-review.md",
     "docs/templates/production-readiness-checklist.md",
+    "docs/templates/worker-dogfood-setup.md",
+    "docs/templates/worker-dogfood-evidence.md",
     "scripts/check-repo.sh",
     "scripts/ci-governance.sh",
     "scripts/ci-tests.sh",
@@ -56,11 +61,23 @@
     "skills/test-driven-development/SKILL.md",
     "skills/code-review-core/SKILL.md",
     "skills/init-capability-evolution/SKILL.md",
-    "skills/webapp-testing/SKILL.md"
+    "skills/webapp-testing/SKILL.md",
+    ".agents/skills/loop-agent/SKILL.md",
+    ".agents/skills/loop-agent/references/command-reference.md",
+    ".agents/skills/ai-engineering-context/SKILL.md",
+    ".agents/skills/verification-before-completion/SKILL.md",
+    ".agents/skills/systematic-debugging/SKILL.md",
+    ".agents/skills/requesting-code-review/SKILL.md",
+    ".agents/skills/codebase-scout/SKILL.md",
+    ".agents/skills/test-driven-development/SKILL.md",
+    ".agents/skills/code-review-core/SKILL.md",
+    ".agents/skills/init-capability-evolution/SKILL.md",
+    ".agents/skills/webapp-testing/SKILL.md"
   ],
   "initSurface": {
     "README.md": "managed-block",
     "AGENTS.md": "managed-block",
+    ".gitignore": "managed-block",
     "harness.json": "generated",
     "docs/README.md": "generated",
     "docs/development-principles.md": "generated",
@@ -69,6 +86,8 @@
     "docs/loop-agent-harness.md": "generated",
     "docs/templates/init-evolution-review.md": "copied",
     "docs/templates/production-readiness-checklist.md": "copied",
+    "docs/templates/worker-dogfood-setup.md": "copied",
+    "docs/templates/worker-dogfood-evidence.md": "copied",
     "scripts/check-repo.sh": "generated",
     "scripts/ci-governance.sh": "generated",
     "scripts/ci-tests.sh": "generated",
@@ -87,7 +106,18 @@
     "skills/test-driven-development/SKILL.md": "copied",
     "skills/code-review-core/SKILL.md": "copied",
     "skills/init-capability-evolution/SKILL.md": "copied",
-    "skills/webapp-testing/SKILL.md": "copied"
+    "skills/webapp-testing/SKILL.md": "copied",
+    ".agents/skills/loop-agent/SKILL.md": "copied",
+    ".agents/skills/loop-agent/references/command-reference.md": "copied",
+    ".agents/skills/ai-engineering-context/SKILL.md": "copied",
+    ".agents/skills/verification-before-completion/SKILL.md": "copied",
+    ".agents/skills/systematic-debugging/SKILL.md": "copied",
+    ".agents/skills/requesting-code-review/SKILL.md": "copied",
+    ".agents/skills/codebase-scout/SKILL.md": "copied",
+    ".agents/skills/test-driven-development/SKILL.md": "copied",
+    ".agents/skills/code-review-core/SKILL.md": "copied",
+    ".agents/skills/init-capability-evolution/SKILL.md": "copied",
+    ".agents/skills/webapp-testing/SKILL.md": "copied"
   },
   "packageExcluded": [
     "docs/progress/20*.md",

package/docs/loop-agent-harness.md

@@ -9,6 +9,7 @@ loop-agent 提供结构化 agent 工作的本地 harness。
 - `.harness/runs/` — one-shot tool run 证据
 - `.harness/cache/` — 本地 runtime 缓存
 - `.harness/live/` — 瞬态 live-session 文件
+- `.task-pool/` — `agent-worker` 的 Task Pool state、batch artifacts、failure handoffs 与 Observe events；它与 `.harness/` 分离，默认不提交
 
 ## Skill 指令
 
@@ -32,6 +33,18 @@ npm 包包含运行与说明 harness 所需的静态能力资产：`bin/`、`dis
 - `loop init`、`loop run`、`loop status`、`loop closeout`
 - `pi-prompt`、`cursor-prompt`
 - `docs audit`、`handoff check`、`spine audit`、`knowledge curate`
+- `agent-worker task validate|explain-profile|retry`、`agent-worker batch run-ready`、`agent-worker report morning`、`agent-worker observe serve|snapshot`
+
+## Worker Retry
+
+失败 Task Pool task 必须显式 retry，不能删除旧 `.task-pool/runs/runs.jsonl` 或复用失败的 `workerRunId`：
+
+```bash
+agent-worker task retry <task-id> --repo <target-repo> --reason "provider configuration corrected"
+agent-worker batch run-ready --feature-dir <feature-dir> --repo <target-repo>
+```
+
+`task retry` 只接受 `Failed` state，保留旧 run 和 failure handoff，并写入 `retryOfWorkerRunId`。下一次 `run-ready` 生成新的 `workerRunId`；`Blocked`、`Done` 或无 state 的 task 必须先由 operator 处理根因，不能盲目重试。
 
 ## 验证 Preset
 

package/docs/reports/README.md

@@ -3,3 +3,7 @@
 本目录存放验证、审计、benchmark 与 closeout 报告。
 
 - [`2026-07-12-production-readiness-hardening.md`](2026-07-12-production-readiness-hardening.md)
+- [`2026-07-10-next-stage-dogfood-runs.md`](2026-07-10-next-stage-dogfood-runs.md) — 发布控制器下的 BE/FE/QA Worker 样本、retry 与 Observe 证据
+- [`2026-07-10-init-evolution-review.md`](2026-07-10-init-evolution-review.md) — Worker 初始化投影演进审查
+- [`2026-07-10-session-events-dogfood.md`](2026-07-10-session-events-dogfood.md) — Pi SDK DAG 新 run 产生非空 `session-events.jsonl` 与 Observe 过程时间线 API 证据
+- [`2026-07-02-repository-analysis.md`](2026-07-02-repository-analysis.md) — 仓库能力基线分析（2026-07-10 再采样：0.5.0 + Observe UI 优化 UI-1～UI-9 / Unreleased）

package/docs/templates/agent-dag.base.json

@@ -60,7 +60,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/docs/templates/agent-dag.final-verification.json

@@ -28,7 +28,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/docs/templates/agent-dag.supervised-implementation.json

@@ -76,7 +76,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/docs/templates/hybrid-dag.json

@@ -56,7 +56,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/docs/templates/worker-dogfood-evidence.md

@@ -0,0 +1,52 @@
+# Worker Dogfood Evidence
+
+## Sample identity
+
+| Field | Value |
+|---|---|
+| Date | |
+| Feature / Task | |
+| Target repo | disposable path or sanitized reference |
+| Controller package/version | |
+| Agent-worker package/version (same npm install) | |
+| Provider/model / override | |
+| Batch ID | |
+| Worker run ID | |
+| Retry of worker run ID | n/a / |
+
+## Baseline
+
+- Baseline commit and `git status`:
+- Existing failing behavior or missing capability:
+- Preflight (`loop-agent --version`, `inspect`, `docs audit`, `check-repo`):
+
+## Run evidence
+
+| Artifact | Path / link | Result |
+|---|---|---|
+| TaskSpec / source-doc copies | | |
+| DAG spec | | |
+| DAG report JSON / Markdown | | |
+| shell verification | | |
+| diff | | |
+| closeout or Failure Handoff | | |
+| morning report | | |
+| Observe snapshot / events | | |
+
+## Acceptance and QA coverage
+
+| Acceptance | Test case(s) | Test file / verification | Result |
+|---|---|---|---|
+| | | | |
+
+## Failure / retry (if applicable)
+
+| Raw failure | Product-line category | Recommended follow-up | Root cause evidence | Retry result |
+|---|---|---|---|---|
+| | | | | |
+
+## Conclusion
+
+- Verdict: pass / fail / blocked
+- Review notes:
+- Follow-up task(s):

package/docs/templates/worker-dogfood-setup.md

@@ -0,0 +1,48 @@
+# Worker Dogfood Setup Template
+
+Use this template to create a disposable target repository for a real `agent-worker` sample. It is deliberately a target-repo recipe, not a substitute for product requirements.
+
+## Preconditions
+
+- Install one published controller version and record it:
+
+  ```bash
+  npm install -g @tea-agent/loop-agent@<version>
+  npm list -g @tea-agent/loop-agent --depth=0
+  loop-agent --version
+  agent-worker --help
+  ```
+
+- Create a clean, disposable target repository. Do not use `npm link`, `npm run dev`, or a workspace controller for a real-evidence run.
+- Put requirement, acceptance, design, test plan, and QA case matrix next to the TaskSpec. `agent-worker` materializes immutable copies into the harness task source before DAG generation.
+
+## Setup checklist
+
+- [ ] Target has a committed baseline and a passing `loop-agent init --profile full --merge` preflight.
+- [ ] Existing test demonstrates the desired behavior is missing or unimplemented.
+- [ ] TaskSpec has narrow `allowed_paths` and explicit `forbidden_paths`.
+- [ ] Task Pool dependencies are either backed by real prior runs or intentionally recorded as pre-existing evidence.
+- [ ] The chosen model/provider and `--pi-model` smoke override, if any, are recorded.
+
+## Execute
+
+```bash
+agent-worker batch run-ready \
+  --feature-dir <feature-dir> \
+  --repo <target-repo> \
+  --limit 1 \
+  --check-repo \
+  [--pi-model <model>]
+
+agent-worker observe snapshot --repo <target-repo> > <evidence-dir>/observe-snapshot.json
+agent-worker report morning --repo <target-repo> --output <evidence-dir>/morning-report.md
+```
+
+For a failed task, do not delete state or alter JSONL evidence. Fix the external/root cause, then make the retry explicit:
+
+```bash
+agent-worker task retry <task-id> --repo <target-repo> --reason "<root cause corrected>"
+agent-worker batch run-ready --feature-dir <feature-dir> --repo <target-repo> --limit 1 --check-repo
+```
+
+The retry run must have a distinct `workerRunId` and retain `retryOfWorkerRunId` in its Task Pool evidence.

package/examples/example-dag.json

@@ -5,7 +5,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/examples/hybrid-loop-agent-dag.json

@@ -57,7 +57,7 @@
     "cursor": {
       "LOW": "composer-2.5",
       "MED": "composer-2.5",
-      "HIGH": "composer-2.5"
+      "HIGH": "gpt-5.5"
     },
     "pi": {
       "LOW": "gpt-5.3-codex-spark",

package/harness.json

@@ -39,7 +39,8 @@
     "execPlansDir": "docs/exec-plans",
     "progressDir": "docs/progress",
     "reportsDir": "docs/reports",
-    "decisionsDir": "docs/decisions"
+    "decisionsDir": "docs/decisions",
+    "taskPoolDir": ".task-pool"
   },
   "scripts": {
     "checkRepo": "scripts/check-repo.sh",
@@ -51,38 +52,13 @@
     "checkInitSurface": "scripts/check-init-surface.sh",
     "ci": "scripts/ci.sh"
   },
-  "models": {
-    "analyze": {
-      "provider": "wizard-local",
-      "model": "gpt-5.3-codex-spark",
-      "timeoutMs": 1800000
-    },
-    "plan": {
-      "provider": "wizard-local",
-      "model": "gpt-5.3-codex-spark",
-      "timeoutMs": 1800000
-    },
-    "implement": {
-      "provider": "wizard-local",
-      "model": "gpt-5.3-codex-spark",
-      "timeoutMs": 1800000
-    },
-    "verify": {
-      "provider": "wizard-local",
-      "model": "gpt-5.3-codex-spark",
-      "timeoutMs": 1800000
-    },
-    "retrospective": {
-      "provider": "wizard-local",
-      "model": "gpt-5.3-codex-spark",
-      "timeoutMs": 1800000
-    }
-  },
   "executors": {
     "pi": {
       "description": "Pi planning, review, diagnosis, and bounded writing when DAG toolProfile=write",
       "defaultModel": "gpt-5.3-codex-spark",
-      "requiresApiKey": ""
+      "LOW": "gpt-5.3-codex-spark",
+      "MED": "glm-5.2",
+      "HIGH": "gpt-5.5"
     },
     "cursor": {
       "description": "Cursor via @cursor/sdk (default model: account default)",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tea-agent/loop-agent",
-  "version": "0.4.0",
+  "version": "0.6.0",
   "type": "module",
   "bin": {
     "loop-agent": "bin/loop-agent.js",
@@ -36,7 +36,7 @@
     "cursor": "node --import tsx/esm src/cli.ts cursor-prompt",
     "pi-prompt": "node --import tsx/esm src/cli.ts pi-prompt",
     "clean": "node -e \"require('node:fs').rmSync('dist',{recursive:true,force:true})\"",
-    "build": "npm run clean && tsc -p tsconfig.build.json",
+    "build": "npm run clean && tsc -p tsconfig.build.json && node -e \"const fs=require('node:fs');const p='dist/worker/observe/static';fs.mkdirSync(p,{recursive:true});fs.cpSync('src/worker/observe/static',p,{recursive:true});\"",
     "prepack": "npm run build",
     "prepublishOnly": "npm run typecheck && npm test && npm run build",
     "lint": "tsc --noEmit",
@@ -54,13 +54,13 @@
   },
   "optionalDependencies": {
     "@cursor/sdk": "^1.0.7",
-    "@earendil-works/pi-coding-agent": "*",
-    "@earendil-works/pi-ai": "*"
+    "@earendil-works/pi-ai": "*",
+    "@earendil-works/pi-coding-agent": "*"
   },
   "devDependencies": {
+    "@types/node": "^24.6.0",
     "tsx": "^4.20.6",
     "typescript": "^5.9.3",
-    "vitest": "^3.2.4",
-    "@types/node": "^24.6.0"
+    "vitest": "^3.2.4"
   }
 }

package/skills/loop-agent/SKILL.md

@@ -35,7 +35,9 @@ Shared loop-agent harness workflow 规则见 `references/harness-policy.md`。Re
 
 ```bash
 loop-agent new-task <task-id> "Task Title"
-# write <repo-root>/.harness/tasks/<task-id>/source/需求.md
+# Prefer immutable original PRD first:
+# loop-agent import-prd <task-id> --file <path-to-original-prd.md>
+# write derived <repo-root>/.harness/tasks/<task-id>/source/需求.md
 # write <repo-root>/.harness/tasks/<task-id>/source/执行约束.md
 
 loop-agent dag run-task <task-id> --profile auto --strict-models --output <temp-dir>/<task-id>-dag.json
@@ -116,8 +118,8 @@ bounded writer 完成后，主会话必须独立复核；命令清单见 `refere
 ## Hard Rules
 
 1. One task = one bounded work chunk.
-2. Source materials are mandatory: `source/需求.md` and `source/执行约束.md`.
-3. Agent DAG is the implementation workflow.
+2. Source materials are mandatory: `source/需求.md` and `source/执行约束.md`. Prefer immutable originals under `source/references/` via `import-prd` or Worker `source_docs`; treat `需求.md` as a derived contract.
+3. Agent DAG is the implementation workflow. Review must three-way check references + derived source + implementation when originals exist.
 4. DAG `pi` executor stays read-only unless the node sets `toolProfile: "write"`; `pi-prompt` / `cursor-prompt` are full-capability one-shot helpers and must be bounded per call.
 5. Pi writer nodes and optional Cursor write execution must be bounded by explicit allowed / forbidden paths.
 6. Completed DAG and one-shot run facts are read-only.

package/skills/loop-agent/references/command-reference.md

@@ -114,6 +114,13 @@ loop-agent new-task <task-id> "Task Title"
 
 创建 `.harness/tasks/<task-id>/`，含 `source/`、`artifacts/`、`logs/` 及初始 state。
 
+### 导入原始 PRD（不可变事实源）
+```bash
+loop-agent import-prd <task-id> --file docs/path/to-prd.md [--name requirement] [--json]
+```
+
+把用户原始 PRD **原样复制** 到 `.harness/tasks/<task-id>/source/references/`，并写入 `source/source-manifest.json`（含 SHA-256）与 `task.json.referenceDocs`。此步骤不调用模型、不改写内容。随后再写派生的 `source/需求.md` 执行契约；冲突时以 `source/references/*` 为准。
+
 ### Task action context / artifact instructions
 ```bash
 loop-agent status <task-id> --json
@@ -137,12 +144,14 @@ loop-agent closeout task <task-id>
 one-shot run evidence 位于 `.harness/runs/{active,completed,failed}/<run-id>/`。`cursor-prompt` 与 Pi `cursor` tool 会创建这类 run evidence；当前 `pi-prompt` 不创建 `.harness/runs/`。active 目录只应保留 live one-shot run，残留或 `.DS_Store` 等系统文件应清理。详见 `one-shot-runs.md`。
 
 ### 运行任何 step 前：准备 source materials
-`new-task` 之后，至少添加 `需求.md`：
+`new-task` 之后，先归档原始 PRD，再写派生执行契约：
 ```bash
+loop-agent import-prd <task-id> --file <path-to-original-prd.md>
+# then write derived contract:
 mkdir -p <repo-root>/.harness/tasks/<task-id>/source
 cat > <repo-root>/.harness/tasks/<task-id>/source/需求.md
 ```
-然后写明需求、目标范围与约束。
+`需求.md` 写目标、范围、非目标与验收标准，并映射回 `source/references/*`；不要让 AI 直接改写原始 PRD。
 
 若 task 有硬约束（仅允许特定文件、禁止改动），另加：
 ```bash
@@ -370,7 +379,7 @@ agent-worker report morning --repo <repo-root> [--batch-run-id <id>] [--output <
 
 - TaskSpec 声明单个任务的业务上下文、`risk_level`、验收与 verify 边界；`risk_level` 被确定性映射到 task complexity。AcceptanceSpec / TaskGraphSpec 声明跨任务验收引用与依赖图，ready queue 决定可运行任务并检测未知依赖/环/文件一致性。
 - 业务 type（`backend-feature`/`frontend-feature`/`qa-testcode` 等）是产品线 profile，不能直接传给 `loop-agent dag run-task --profile`；Worker 会映射到 `auto`/`minimal`/`standard`/`reviewed`/`supervised`。
-- materializer 把 TaskSpec 物化为 `.harness/tasks/<task-id>` 后，Worker 调用 `dag run-task` / `dag validate` / `run-dag` / `dag report`。成功路径走 `promote-run` + `closeout task`；失败路径收集 `dag doctor` / `dag closeout-draft` evidence，写入目标 repo 的 `.task-pool/failure-handoffs/`（不写 `.harness/`，因为 `dag closeout-draft` 会拒绝在 completed facts 之外写入）。
+- materializer 把 TaskSpec 物化为 `.harness/tasks/<task-id>`：`source_docs` 原样进入 `source/references/`，派生 `需求.md` 带权威声明、Source Docs/hash 追溯，以及 `acceptance_refs` 短摘要；随后 Worker 调用 `dag run-task` / `dag validate` / `run-dag` / `dag report`。review 节点须对照 references + 派生契约 + 实现。成功路径走 `promote-run` + `closeout task`；失败路径收集 `dag doctor` / `dag closeout-draft` evidence，写入目标 repo 的 `.task-pool/failure-handoffs/`（不写 `.harness/`，因为 `dag closeout-draft` 会拒绝在 completed facts 之外写入）。
 - Worker runtime state 落在目标 repo 的 `.task-pool/`（artifacts、JSONL/state、晨报、failure handoffs），与 `.harness/` 分离；`.task-pool/` 默认被 `.gitignore` 忽略。
 - preflight 在 `new-task` 前跑 `loop-agent --version`、`inspect`、`docs-audit`、`git status --short --branch`，可选 `--check-repo`。一次夜间批处理期间不升级控制器，记录实际 `loop-agent` 版本。
 - `batch run-ready` 默认在 stderr 输出人类可读进度（批次起止、每个 task 的阶段与耗时、report 决策），stdout 只保留最终 JSON，便于管道取用；加 `--quiet` 可关闭进度。