tuner-testkit 4.3.0__tar.gz → 5.0.0__tar.gz

{tuner_testkit-4.3.0/tuner_testkit.egg-info → tuner_testkit-5.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tuner-testkit
-Version: 4.3.0
+Version: 5.0.0
 Summary: Shared test runtime, recorders, and DNA overlay for SUT test repos
 Author: tuner-testkit contributors
 License-Expression: MIT
@@ -167,6 +167,10 @@ git add -A && git commit
 
 `tuner-dna check` 核对项目树 DNA 是否与**当前 PATH 上那份** kit 一致。建议先 `uv tool upgrade`，再 `sync`，避免 CLI 还是旧包、`.venv` 已是新包。
 
+**其他 AI IDE**：DNA 以 Cursor 布局为源，`tuner-dna sync --ide claude,agents`（或 `all`）把 skills 渲染到
+`.claude/skills/` / `.agents/skills/`、subagents 到 `.claude/agents/`，并生成 `CLAUDE.md`（`@AGENTS.md` + always-on 规则内联 + 路径规则清单）。
+`tuner-dna targets` 列出各目标落点；hooks.json 不跨工具翻译。目录约定随各工具文档变动，用前核对。
+
 新仓第一次仍用第 2 节的 `tuner-init scaffold`（会写依赖并做一次 `tuner-dna sync`），不是每次发版都重新 scaffold。
 
 `[phone-ui]` 为预留空 extra，4.0 未实现。
@@ -210,4 +214,8 @@ uv sync --extra all
 
 `uv sync --extra dev` 与 `--extra all` 等价（`dev` 只是引用 `[all]`）。
 
-4.3.0 发行说明见 [`docs/changelog/TEMPLATE-4.3.0.md`](docs/changelog/TEMPLATE-4.3.0.md)；4.2.0 见 [`TEMPLATE-4.2.0.md`](docs/changelog/TEMPLATE-4.2.0.md)；3.x → 4.x 迁移见 [`TEMPLATE-4.0.0.md`](docs/changelog/TEMPLATE-4.0.0.md)。平台能力地图见 [`INDEX.md`](INDEX.md)。
+本仓是**平台仓**，不是 workspace。`dogfood/` 是用 `tuner-init scaffold dogfood --no-ai` 生成的 workspace 示范
+（被测系统 = 平台本身，uv workspace 成员，共用 `.venv`）：路线图在 `dogfood/assets/domain-notes/platform/roadmap.md`，
+验收活文档在 `dogfood/tests/features/`。见 [`dogfood/README.md`](dogfood/README.md)。
+
+5.0.0 发行说明（Plane Job 协议退场、`tuner-workspace`）见 [`docs/changelog/TEMPLATE-5.0.0.md`](docs/changelog/TEMPLATE-5.0.0.md)；4.3.0 见 [`TEMPLATE-4.3.0.md`](docs/changelog/TEMPLATE-4.3.0.md)；3.x → 4.x 迁移见 [`TEMPLATE-4.0.0.md`](docs/changelog/TEMPLATE-4.0.0.md)。平台能力地图见 [`INDEX.md`](INDEX.md)。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/README.md

@@ -122,6 +122,10 @@ git add -A && git commit
 
 `tuner-dna check` 核对项目树 DNA 是否与**当前 PATH 上那份** kit 一致。建议先 `uv tool upgrade`，再 `sync`，避免 CLI 还是旧包、`.venv` 已是新包。
 
+**其他 AI IDE**：DNA 以 Cursor 布局为源，`tuner-dna sync --ide claude,agents`（或 `all`）把 skills 渲染到
+`.claude/skills/` / `.agents/skills/`、subagents 到 `.claude/agents/`，并生成 `CLAUDE.md`（`@AGENTS.md` + always-on 规则内联 + 路径规则清单）。
+`tuner-dna targets` 列出各目标落点；hooks.json 不跨工具翻译。目录约定随各工具文档变动，用前核对。
+
 新仓第一次仍用第 2 节的 `tuner-init scaffold`（会写依赖并做一次 `tuner-dna sync`），不是每次发版都重新 scaffold。
 
 `[phone-ui]` 为预留空 extra，4.0 未实现。
@@ -165,4 +169,8 @@ uv sync --extra all
 
 `uv sync --extra dev` 与 `--extra all` 等价（`dev` 只是引用 `[all]`）。
 
-4.3.0 发行说明见 [`docs/changelog/TEMPLATE-4.3.0.md`](docs/changelog/TEMPLATE-4.3.0.md)；4.2.0 见 [`TEMPLATE-4.2.0.md`](docs/changelog/TEMPLATE-4.2.0.md)；3.x → 4.x 迁移见 [`TEMPLATE-4.0.0.md`](docs/changelog/TEMPLATE-4.0.0.md)。平台能力地图见 [`INDEX.md`](INDEX.md)。
+本仓是**平台仓**，不是 workspace。`dogfood/` 是用 `tuner-init scaffold dogfood --no-ai` 生成的 workspace 示范
+（被测系统 = 平台本身，uv workspace 成员，共用 `.venv`）：路线图在 `dogfood/assets/domain-notes/platform/roadmap.md`，
+验收活文档在 `dogfood/tests/features/`。见 [`dogfood/README.md`](dogfood/README.md)。
+
+5.0.0 发行说明（Plane Job 协议退场、`tuner-workspace`）见 [`docs/changelog/TEMPLATE-5.0.0.md`](docs/changelog/TEMPLATE-5.0.0.md)；4.3.0 见 [`TEMPLATE-4.3.0.md`](docs/changelog/TEMPLATE-4.3.0.md)；3.x → 4.x 迁移见 [`TEMPLATE-4.0.0.md`](docs/changelog/TEMPLATE-4.0.0.md)。平台能力地图见 [`INDEX.md`](INDEX.md)。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "tuner-testkit"
-version = "4.3.0"
+version = "5.0.0"
 description = "Shared test runtime, recorders, and DNA overlay for SUT test repos"
 readme = "README.md"
 license = "MIT"
@@ -69,7 +69,7 @@ tuner-mock-server = "tuner_testkit.apps.mock_server.cli:main"
 tuner-dna = "tuner_testkit.apps.dna.cli:main"
 tuner-init = "tuner_testkit.apps.init_repo.__main__:main"
 tuner-index-ai = "tuner_testkit.apps.index_ai.__main__:main"
-tuner-index-platform = "tuner_testkit.apps.index_platform.__main__:main"
+tuner-workspace = "tuner_testkit.workspace.cli:main"
 tuner-fake = "tuner_testkit.fake.cli:main"
 tuner-page-test = "tuner_testkit.page_test.__main__:main"
 tuner-action-words = "tuner_testkit.action_words.__main__:main"
@@ -90,6 +90,10 @@ include = ["tuner_testkit*"]
 build_py = "tuner_testkit.apps.dna.build.BuildPyWithDna"
 sdist = "tuner_testkit.apps.dna.build.SdistWithDna"
 
+[tool.uv.workspace]
+# dogfood/ is a workspace scaffolded by `tuner-init` from this repo (SUT = the platform itself).
+members = ["dogfood"]
+
 [tool.tuner-testkit]
 page_objects = "packages/page_objects"
 api_objects = "packages/api_objects"

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/__init__.py

@@ -2,4 +2,4 @@
 
 from __future__ import annotations
 
-__version__ = "4.3.0"
+__version__ = "4.4.0"

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/api_mock/store.py

@@ -6,9 +6,8 @@ Two layers:
 * **overrides** — whatever the admin API changed since startup.
 
 Overrides win on read and stay in memory until :meth:`MockStore.persist` writes
-them out. That split is deliberate: the Plane runner must be able to redefine a
-response without the server writing into the checked-out repo (see
-``tuner_testkit/apps/index_platform/README.md`` — catalog jobs do not write git).
+them out. That split is deliberate: a test platform or the workbench must be able
+to redefine a response without the server writing into the checked-out repo.
 """
 from __future__ import annotations
 

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/build.py

@@ -31,8 +31,19 @@ def _payload_ready(root: Path | None = None) -> bool:
 
 
 def _bundle() -> None:
+    """Refresh ``dna_payload`` from the template tree when building a wheel.
+
+    A leftover ``dna_payload/`` from an older build must not short-circuit:
+    that shipped stale DNA (4.3.0 payload inside a 4.4.0 wheel).
+    """
     _ensure_src_on_path()
-    if _payload_ready():
+    from tuner_testkit.project import template_source_root
+
+    if template_source_root() is None:
+        if not _payload_ready():
+            raise RuntimeError(
+                "no DNA payload in this install and no template checkout to bundle from"
+            )
         return
 
     from tuner_testkit.apps.dna.cli import cmd_bundle

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/cli.py

@@ -3,7 +3,6 @@
 from __future__ import annotations
 
 import argparse
-import hashlib
 import json
 import shutil
 import sys
@@ -16,43 +15,39 @@ from tuner_testkit.apps.dna.source import (
     iter_dna_files,
     payload_package_dir,
 )
+from tuner_testkit.apps.dna.targets import (
+    TARGET_LAYOUTS,
+    diff_file_map,
+    parse_targets,
+    render_target,
+    write_file_map,
+)
 from tuner_testkit.project import ProjectRootError, project_root
 
 
-def _hash_bytes(data: bytes) -> str:
-    return hashlib.sha256(data).hexdigest()
-
-
 def _target_root(explicit: str | None) -> Path:
     if explicit:
         return Path(explicit).expanduser().resolve()
     return project_root()
 
 
+def _targets(args: argparse.Namespace) -> list[str]:
+    return parse_targets(getattr(args, "ide", None))
+
+
 def cmd_sync(args: argparse.Namespace) -> int:
     try:
         source = dna_source_root()
         dest = _target_root(args.target)
-    except (DnaSourceError, ProjectRootError) as exc:
+        targets = _targets(args)
+    except (DnaSourceError, ProjectRootError, ValueError) as exc:
         print(str(exc), file=sys.stderr)
         return 2
 
-    copied = 0
-    skipped = 0
-    for rel, src in iter_dna_files(source):
-        dst = dest / rel
-        if dst.exists() and not args.overwrite:
-            if dst.is_file() and src.read_bytes() == dst.read_bytes():
-                skipped += 1
-                continue
-            if dst.is_file() and not args.overwrite:
-                skipped += 1
-                continue
-        dst.parent.mkdir(parents=True, exist_ok=True)
-        shutil.copy2(src, dst)
-        copied += 1
-        print(f"copy {rel}")
-    print(f"dna sync: copied={copied} skipped={skipped} dest={dest}")
+    for target in targets:
+        files = render_target(source, target)
+        copied, skipped = write_file_map(dest, files, overwrite=args.overwrite)
+        print(f"dna sync[{target}]: copied={copied} skipped={skipped} dest={dest}")
     marker = dest / ".tuner-dna-version"
     marker.write_text(_kit_version() + "\n", encoding="utf-8")
     return 0
@@ -62,28 +57,33 @@ def cmd_check(args: argparse.Namespace) -> int:
     try:
         source = dna_source_root()
         dest = _target_root(args.target)
-    except (DnaSourceError, ProjectRootError) as exc:
+        targets = _targets(args)
+    except (DnaSourceError, ProjectRootError, ValueError) as exc:
         print(str(exc), file=sys.stderr)
         return 2
 
-    missing: list[str] = []
-    changed: list[str] = []
-    for rel, src in iter_dna_files(source):
-        dst = dest / rel
-        if not dst.is_file():
-            missing.append(rel)
+    drifted = False
+    for target in targets:
+        missing, changed = diff_file_map(dest, render_target(source, target))
+        if not missing and not changed:
+            print(f"DNA[{target}] matches installed kit payload.")
             continue
-        if _hash_bytes(src.read_bytes()) != _hash_bytes(dst.read_bytes()):
-            changed.append(rel)
-    if not missing and not changed:
-        print("DNA matches installed kit payload.")
-        return 0
-    for rel in missing:
-        print(f"  missing: {rel}", file=sys.stderr)
-    for rel in changed:
-        print(f"  changed: {rel}", file=sys.stderr)
-    print("Run `tuner-dna sync` to refresh platform DNA.", file=sys.stderr)
-    return 1
+        drifted = True
+        for rel in missing:
+            print(f"  [{target}] missing: {rel}", file=sys.stderr)
+        for rel in changed:
+            print(f"  [{target}] changed: {rel}", file=sys.stderr)
+    if drifted:
+        print("Run `tuner-dna sync --overwrite [--ide <targets>]` to refresh platform DNA.", file=sys.stderr)
+        return 1
+    return 0
+
+
+def cmd_targets(_args: argparse.Namespace) -> int:
+    for name, layout in TARGET_LAYOUTS.items():
+        print(f"{name:8s} skills={layout.skills_dir or '-':16s} agents={layout.agents_dir or '-':16s} "
+              f"instructions={layout.instructions_file or '-':10s} {layout.note}")
+    return 0
 
 
 def cmd_bundle(_args: argparse.Namespace) -> int:
@@ -123,19 +123,25 @@ def _kit_version() -> str:
 def main(argv: list[str] | None = None) -> int:
     parser = argparse.ArgumentParser(
         prog="tuner-dna",
-        description="Sync Cursor/git DNA from tuner-testkit into a project repo.",
+        description="Sync AI-IDE / git DNA from tuner-testkit into a project repo (Cursor by default; --ide for others).",
     )
     sub = parser.add_subparsers(dest="cmd", required=True)
+    ide_help = "comma-separated IDE targets: cursor (default), claude, agents, codex, all"
 
-    p_sync = sub.add_parser("sync", help="copy DNA files into the project")
+    p_sync = sub.add_parser("sync", help="copy / render DNA files into the project")
     p_sync.add_argument("--target", default=None, help="project root (default: cwd project)")
     p_sync.add_argument("--overwrite", action="store_true", help="overwrite differing files")
+    p_sync.add_argument("--ide", default=None, help=ide_help)
     p_sync.set_defaults(func=cmd_sync)
 
     p_check = sub.add_parser("check", help="report DNA drift vs the kit payload")
     p_check.add_argument("--target", default=None)
+    p_check.add_argument("--ide", default=None, help=ide_help)
     p_check.set_defaults(func=cmd_check)
 
+    p_targets = sub.add_parser("targets", help="list IDE targets and where each renders")
+    p_targets.set_defaults(func=cmd_targets)
+
     p_bundle = sub.add_parser("bundle", help="copy DNA into the wheel payload directory")
     p_bundle.set_defaults(func=cmd_bundle)
 

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/agents/bdd-asset-pipeline.md

@@ -1,6 +1,6 @@
 ---
 name: bdd-asset-pipeline
-version: 2.0.0
+version: 2.0.1
 description: Lower-half of SUT self-learning — Feature Sets + design knowledge → Reuse Analysis → Freeze API Objects → Maintain Page Objects → Maintain Steps → behave regression. Use only for behave/BDD automation assets. Do NOT use when the user explicitly requests pytest. Do NOT use this agent as the explore-first entry; start from docs/spec/sut-self-learning.md instead.
 ---
 
@@ -104,3 +104,5 @@ Before writing Then-assertions, **read `assets/design/**`**:
 - `behave --stage ui` succeeds for UI features that exist.
 - `behave --stage api` succeeds for API features that exist (or `--dry-run`
   with documented TODO).
+- Task log written to `artifacts/inbox/` per `docs/spec/agent-task-log.md`
+  (no secrets).

tuner_testkit-5.0.0/tuner_testkit/apps/dna/dna_payload/.cursor/agents/sut-self-learning.md

@@ -0,0 +1,72 @@
+---
+name: sut-self-learning
+version: 1.0.0
+description: Orchestrate explore-first SUT self-learning — accounts → Playwright MCP explore → evidence → freeze API/page objects → design knowledge → index, with an artifacts/inbox task log. Hand off to bdd-asset-pipeline only when Feature Sets are in scope. Do NOT use when the user explicitly requests pytest-only work.
+---
+
+# SUT self-learning orchestrator (upper half + optional handoff)
+
+## Purpose
+
+串起「学习被测 Web 系统并生长测试资产」的**上半程**。规格
+`docs/spec/sut-self-learning.md`。本文件是可执行编排；规格不是本 agent 的替代。
+
+**下半程**（已有 feature set → steps / behave）交给
+`.cursor/agents/bdd-asset-pipeline.md`，不要在本 agent 里硬造 Gherkin，除非用户明确要验收自动化。
+
+## When to use
+
+- 工程师说「探索 SUT / 学习某某模块 / 生成测试资产 / 再跑一轮自学习」。
+- 下游仓已有 `tuner-init` 骨架，SUT 可访问。
+
+## Inputs
+
+1. `INDEX.md` + 若存在 `INDEX.project.md`（已有切片、勿重复冻登录接口）。
+2. **账号**（只读，禁止拷进仓库/evidence/inbox）：
+   1. `data/sut-accounts.local.yaml`（gitignore，首选）
+   2. 否则 `config/env_local.py` 的 `TEST_ACCOUNT`
+   3. 都没有则停，让工程师按 `data/sut-accounts.example.yaml` 补本机文件
+3. 垂直切片（一个模块，不是全站爬）。
+4. 可选：SUT 源码根（如 Plane `apps/api`）——有则跑 `sut-source-to-design`。
+
+## Skills / agents（顺序）
+
+| 步 | 组件 | 产出 |
+| --- | --- | --- |
+| 0 | 读账号 + `explore-safety` | 登录参数（内存） |
+| 1 | `dump-ddl`（缺表才跑，禁 `--all`） | `assets/ddl/<ds>/` |
+| 2 | `explore-sut` | `artifacts/evidence/<run_id>/` + `assets/explore/` |
+| 3 | `analyze-mcp-network`（`tuner-evidence routes`） | 路由表 JSON |
+| 4 | `reuse-analysis` | 复用 vs 新建 |
+| 5 | `freeze-api-objects` | `packages/api_objects/`（跳过 sign-in） |
+| 6 | `maintain-page-objects` | `packages/page_objects/` |
+| 7 | `derive-design-knowledge`；有源码则 **`sut-source-to-design`** | `assets/design/` |
+| 8 | `maintain-index` | `INDEX.project.md` 或 `INDEX.md` |
+| 9 | **本 agent** 写 inbox | `artifacts/inbox/<utc>-sut-self-learning-<slice>.md` |
+| 10 | 仅当用户要 `.feature` | `derive-feature-sets` → **`bdd-asset-pipeline`** |
+
+每步先 **Read** 对应 `SKILL.md` / agent 正文再执行。不要跳过 evidence 落盘。
+
+## Inbox（强制）
+
+遵循 `docs/spec/agent-task-log.md`。SUT 自学习日志 **必须**含：
+
+- `## Pages explored`（url pattern / title / 按钮）
+- `## Assets added this run`（本任务新增或更新的 api/page/explore/design/ddl）
+- `run_id`；`account_id`（无密码）
+- 若这是第 N 轮 dogfood：`## Loop delta` 对比上一轮 inbox / 平台能力
+
+## Hard rules
+
+- 活体 SUT：`explore-safety`（throwaway 数据；不问不准删）。
+- 密钥不入库、不进 evidence、不进 inbox。
+- Web cookie session 的 path 占位用 `{workspace_slug}` 等，调用走 `set_path`。
+- 页面资产 app 名对齐现有 `packages/page_objects/<app>/`，勿与装饰器模块重名。
+- 未走 Gate 1 证据不得写 api/page/design。
+
+## Done
+
+- evidence 在磁盘且已脱敏。
+- 切片的 explore +（若有写操作）design 已引用 `run_id`。
+- 新路由已冻或书面说明跳过原因。
+- inbox 已落盘，页面清单与资产增量齐全。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/agents/sut-source-to-design.md

@@ -1,6 +1,6 @@
 ---
 name: sut-source-to-design
-version: 1.0.0
+version: 1.0.1
 description: Walk SUT backend source (Django models/views/serializers, or equivalent) plus assets/ddl to draft assets/design/<app>/<route-slug>.md. Use when the engineer points at a source tree. Do not invent tables. Complements derive-design-knowledge; this agent is the source-channel playbook.
 ---
 
@@ -31,3 +31,5 @@ Highest-quality `writes` / `noise` / `reads` often come from **models.save / vie
 - `confidence: inferred` for source-only; `observed` only with DB/log evidence.
 - No secrets.
 - Do not generate action words here.
+- Write `artifacts/inbox/<utc>-sut-source-to-design-<slice>.md` per
+  `docs/spec/agent-task-log.md` (outputs = design files; no secrets).

tuner_testkit-5.0.0/tuner_testkit/apps/dna/dna_payload/.cursor/rules/agent-task-log.mdc

@@ -0,0 +1,22 @@
+---
+description: Every agent run must leave a task log under artifacts/inbox/ (no secrets).
+alwaysApply: true
+version: 1.0.0
+---
+
+# Agent task log — `artifacts/inbox/`
+
+每次按 `.cursor/agents/*.md` 编排执行（含 SUT 自学习），结束前必须写一份任务记录。
+细则见 `docs/spec/agent-task-log.md`。
+
+## Must
+
+- 路径：`artifacts/inbox/<UTC>-<agent>-<slice>.md`（例 `20260914T031200Z-sut-self-learning-cycles.md`）。
+- 必填：`agent`、意图、输入、产出文件列表、下一步。
+- **SUT 自学习**额外必填：本次探索的页面（url pattern + 标题）、本次新增/更新的测试资产路径。
+- 只写 `account_id` / workspace slug，**禁止**密码、token、Cookie。
+
+## Must not
+
+- 用聊天摘要代替 inbox 落盘。
+- 把 inbox 当 evidence（网络 dump 仍走 `artifacts/evidence/<run_id>/`）。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/rules/apps-authoring.mdc

@@ -2,7 +2,7 @@
 description: SUT apps/ and kit tuner_testkit.apps tools must be standalone-runnable and reuse tuner_testkit plus this repo's packages business assets; no reinventing wheels, no secrets, not imported by tests
 globs: apps/**,tuner_testkit/apps/**
 alwaysApply: false
-version: 1.4.0
+version: 2.0.0
 ---
 
 # apps — 创建期约束
@@ -30,18 +30,20 @@ version: 1.4.0
 - 不得被 `tests/` import（复用逻辑下沉到 `packages/` 业务资产或 `tuner_testkit`）。
 - 不得写入 token / cookie / 密码 / Authorization / session 等敏感值。
 - 破坏性操作无开关、默认非 dry-run/只读 → 禁止。
-- 不得为了「被 Plane 看见」而给本地维护工具加 `@plane_app`（dump_ddl / recorder /
-  api_recorder / page_recorder / init_repo / index_ai 保持本机 CLI，产物进 git）。
-
-## Plane 可运行 vs 本地维护
-
-- **本地维护**（工程师跑完再 commit）：不必 `plane.py`，Plane Runner 不会执行。
-- **Plane Job**：必须提供 import-safe 的 `apps/<name>/plane.py`，用
-  `tuner_testkit.apps._shared.plane_app.@plane_app` / `register_plane_app` 登记；SUT CLI 为
-  `python -m apps.<name>`。领域逻辑在 `packages/`，apps 只做协议适配。
-  造数 / API / UI 动作不要再做成伪 app；给 word 加 `@plane_db_seed` 等即可。
-- 发现器只 import 存在 `plane.py` 的包；`cli.py` 可以较重，但 `plane.py` 禁止
-  顶层 import DB / mitmproxy。
+- 造数 / API / UI 动作不得做成伪 app：写成 action word（`@register` 后自动进目录与工作台）。
+
+## 工作台可见 vs 本地维护
+
+- **工作台可见**（团队成员要在 `tuner-workbench` 里填参运行）：提供 import-safe 的
+  `apps/<name>/tool.py`，用 `tuner_testkit.tools.tool(tool_id=..., name=..., module="apps.<name>", group=...,
+  destructive=..., timeout=...)` 装饰 `build_parser`；`tuner-workspace catalog` 据此导出
+  `params_schema` / `argv_plan`。每个参数必须有 `help`（即表单 label）；破坏性工具声明 `destructive=True`
+  且 CLI 默认 dry-run。
+- **本地维护**（工程师跑完再 commit，产物进 git；如 dump_ddl / recorder / init_repo / index_ai）：
+  不写 `tool.py`，或写了但 `visibility="local"`。
+- 发现器只 import `apps/*/tool.py` 与 kit 的 `tuner_testkit/apps/*/tool.py`；`cli.py` 可以较重，
+  但 `tool.py` 及其 import 的模块禁止顶层 import DB 驱动 / Playwright / mitmproxy。
+- 需要机器可读输出时用 `--json` envelope（`status / outputs[] / artifacts[] / log_path`），见规范 §7。
 
 ## 相关
 

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/rules/bdd-asset-layering.mdc

@@ -2,7 +2,7 @@
 description: Enforce layered boundaries between feature/steps/action_words/page_objects/api_objects and tuner_testkit runtime; no selector or request leakage across layers
 globs: **/*.py
 alwaysApply: false
-version: 1.3.1
+version: 1.4.0
 ---
 
 # Asset Layering — boundary rules per layer
@@ -43,8 +43,8 @@ Each layer owns one responsibility. Leakage across layers (e.g. selectors in ste
 - **No secrets in repo**, ever: `Authorization`, `Cookie`, `Set-Cookie`, `*token*`, `*secret*`, `*password*`, `*session*`. Also applies to captures under `artifacts/evidence/`.
 - **Evidence-driven edits**: Page/API object changes must be backed by Playwright MCP snapshots, `tuner-recorder` / `tuner-api-recorder` captures (`python -m tuner_testkit.apps.recorder` / `api_recorder`), or run_summary findings; avoid locator/path guesses.
 - **`tests/**` must not import `apps.*`**. Action words for BDD/pytest stay in
-  `tuner_testkit.action_words`. Opt into Plane with `@plane_db_seed` (etc.) from
-  `tuner_testkit.action_words.plane`; do not wrap them as `apps.*`.
+  `tuner_testkit.action_words`; every `@register`ed word is already listed by
+  `tuner-workspace catalog` and runnable from the workbench — do not wrap them as `apps.*`.
 
 ## Examples
 

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/rules/commit-convention.mdc

@@ -1,7 +1,7 @@
 ---
 description: Commits must follow Conventional Commits with a layer-aligned scope; the declared scope must cover the layers actually changed. Single source of truth is docs/spec/commit-convention.md.
 alwaysApply: true
-version: 1.0.0
+version: 1.1.0
 ---
 
 # 提交规范（Conventional Commits + 六层 scope）
@@ -12,7 +12,8 @@ version: 1.0.0
 - `type` ∈ `feat|fix|refactor|docs|test|perf|chore|build|ci|style|revert`。
 - `scope` 用**对齐六层/资产类别**的值：`assets / apps / packages / api_objects /
   page_objects / action_words / tests / docs / rules / skills / agents / hooks /
-  init_repo / index`（多层用逗号分隔，如 `feat(apps,packages): ...`）。
+  init_repo / index / dogfood`（多层用逗号分隔，如 `feat(apps,packages): ...`；
+  `dogfood/**` 一律用 `dogfood`）。
 - **scope 必须覆盖本次真实改动涉及的层**（`packages` 是三个子资产层的上位 scope）。
 
 一眼可读优先：让人从提交信息就看出"更新了 assets、新增了 apps、录制了新的 api_objects"等。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/rules/index-hygiene.mdc

@@ -1,13 +1,14 @@
 ---
 description: Keep the repo-level INDEX.md current and require managed assets to leave a changelog trail; index is the discovery entry for humans and the LLM
-globs: assets/**,packages/api_objects/**,packages/page_objects/**,apps/**,tuner_testkit/apps/**
+globs: assets/**,packages/api_objects/**,packages/page_objects/**,apps/**,tuner_testkit/apps/**,dogfood/assets/**,dogfood/packages/**,dogfood/apps/**,dogfood/INDEX.project.md
 alwaysApply: false
-version: 1.1.3
+version: 1.3.0
 ---
 
 # index-hygiene — 索引与留痕纪律
 
 根 `INDEX.md` 是**平台**能力地图（发现入口）。若存在 `INDEX.project.md`，那是试验田 / 下游项目仓的**业务资产**地图；两份索引都必须与各自职责范围内的仓库状态一致。
+平台仓内的 `dogfood/` 是用 scaffold 生成的 workspace，它的业务地图是 `dogfood/INDEX.project.md`。
 
 ## 留痕（changelog trail）
 
@@ -21,7 +22,9 @@ version: 1.1.3
 
 ## 增量维护（省 token）
 
-- 更新 `INDEX.md` 时**优先读各区 `CHANGELOG` 的 delta**，据此补条目；
+- 索引里 `<!-- auto:begin:<zone> -->` 围栏内的区块由 `tuner-workspace index render` 从
+  `artifacts/catalogs/workspace.json` 确定性渲染，**禁止手写**；CI 跑 `tuner-workspace index check`。
+- 围栏外的人工区更新 `INDEX.md` 时**优先读各区 `CHANGELOG` 的 delta**，据此补条目；
   **不要为维护索引而全量读取 `assets/ddl/*.sql` 正文**。
 - AI 组件区（`.cursor/**`）不手写进 `INDEX.md`：由 `python -m tuner_testkit.apps.index_ai`
   确定性生成 `.cursor/REGISTRY.md`，`INDEX.md` 只链接过去。

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/skills/explore-sut/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: explore-sut
-version: 1.1.0
+version: 1.2.0
 description: Explore-first Playwright MCP session against a live SUT. Produces sanitized artifacts/evidence/<run_id>/ plus assets/explore capability notes. Use when learning a web SUT, mapping pages/buttons, or gathering evidence before any .feature exists. Do NOT use when the user already has a feature to implement — use run-feature-playwright-mcp instead.
 ---
 
@@ -23,6 +23,10 @@ description: Explore-first Playwright MCP session against a live SUT. Produces s
 ## Steps
 
 1. Read `INDEX.md` and `INDEX.project.md` if present. Note base URLs from `config`.
+   Load login from **`data/sut-accounts.local.yaml`** if it exists; else
+   `config/env_local.py` `TEST_ACCOUNT`. Copy `data/sut-accounts.example.yaml`
+   when neither exists and **stop**. Never write the password into evidence,
+   explore notes, inbox, or git.
 2. Pick a **vertical slice** (e.g. workspace → project → issue), not a full-site crawl.
 3. Navigate with Playwright MCP. After each meaningful action, dump network + snapshot.
 4. Persist with **`tuner-evidence persist`** (do not leave captures only in chat):

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/skills/maintain-index/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: maintain-index
 description: Updates the repo-level INDEX.md (knowledge/capability map) incrementally by reading changelog deltas, not full asset bodies. Use after new assets/tools/objects are added, or when INDEX.md drifts from reality.
-version: 1.1.2
+version: 1.2.0
 ---
 
 # Maintain INDEX.md（增量、省 token）
@@ -13,6 +13,19 @@ version: 1.1.2
 - **写范围**：`INDEX.md`；若 `INDEX.project.md` 存在则也可改它。AI 组件区不手写（见下）。
 - **分流**：平台工具 / 公共 `packages` / AI 组件链接 → `INDEX.md`。业务 DDL / SQL / 用例 / explore / design / api_objects / page_objects / features → 有 `INDEX.project.md` 就写那里，不要把 SUT 明细填进 `INDEX.md`。
 
+## 先跑确定性渲染，再做人工区
+
+索引文件里被 `<!-- auto:begin:<zone> -->` … `<!-- auto:end:<zone> -->` 围起来的区块**不要手写**：
+
+```bash
+uv run tuner-workspace catalog          # 刷新 artifacts/catalogs/workspace.json
+uv run tuner-workspace index render     # 渲染 INDEX.project.md（无则 INDEX.md）的 auto 围栏
+uv run tuner-workspace index check      # CI：自动区过期则 exit 1
+```
+
+zone：`ddl` / `sql` / `assets` / `api_objects` / `page_objects` / `action_words` / `apps` / `features`。
+本 skill 只负责围栏**之外**的人工区（用途一句话、confidence 采信说明、平台工具的交接文档链接等）。
+
 ## 核心原则：读 delta，不全量读
 
 - **自动区**（`assets/ddl/`、`packages/api_objects/`、`packages/page_objects/` 等）：**只读各区 `CHANGELOG` 的尾部增量**

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/.cursor/skills/maintain-page-objects/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: maintain-page-objects
-version: 2.0.2
+version: 2.1.0
 description: Creates or updates page assets under packages/page_objects/ on the tuner_testkit.page_test runtime (PageModel element table + declarative flows, or BasePage escape hatch) following page-objects-syntax.md. Use when UI steps fail due to locators, new UI flows are needed, selector leakage must be removed from steps, or doctor reports locator rot.
 ---
 
@@ -50,7 +50,7 @@ description: Creates or updates page assets under packages/page_objects/ on the
 - 每个 flow 带 `description` 与 `example_params`（让 `run --example` 可一键跑）。
 - 页面级 `asserts` / `extracts` 只放**页面自身的稳定契约**（`open()` 时执行）；
   flow 级断言写在该 flow 的 steps 里。
-- 需要上 Plane Formulation 时加 `@plane_pageobject`。
+- 不需要任何装饰器就会出现在 `tuner-workspace catalog` 与工作台的只读页面清单里（静态扫描 `page_id`）。
 
 ## Quick validation（必须跑）
 

{tuner_testkit-4.3.0 → tuner_testkit-5.0.0}/tuner_testkit/apps/dna/dna_payload/AGENTS.md

@@ -1,4 +1,4 @@
-<!-- version: 1.3.0 -->
+<!-- version: 1.5.0 -->
 # AGENTS — 平台总入口（轻量 DM）
 
 > 本文件是"repo as a platform"的**稳定世界观**：只写不常变的分层结构与角色。
@@ -21,7 +21,7 @@
 | 测试层 | `tests/` | 自动化测试代码（behave `features/` + `pytest/`）。 | behave / pytest |
 | 文档层 | `docs/` | 仓库使用说明与规范（`docs/spec/**`）。 | 只读 |
 
-配套：`config/`（环境/数据源）、`.cursor/`（AI 组件，见下）、`artifacts/`·`logs/`（运行产出；MCP evidence 落 `artifacts/evidence/<run_id>/`）。
+配套：`config/`（环境/数据源）、`.cursor/`（AI 组件，见下）、`artifacts/`·`logs/`（运行产出，规范 `docs/spec/artifacts-layout.md`：MCP evidence 落 `artifacts/evidence/<run_id>/`；agent 任务记录落 `artifacts/inbox/`；工具运行落 `artifacts/runs/<run_id>/`；回归报告落 `artifacts/reports/<run_id>/`，每个 run 目录带 `manifest.json`）。SUT 登录本机文件：`data/sut-accounts.local.yaml`（gitignore）。
 
 ## 2. 三角色
 
@@ -43,7 +43,7 @@
 ```mermaid
 flowchart TD
   Start[任务到达] --> Q{任务类型?}
-  Q -->|"学习被测系统 / 探索式测试"| Learn["SUT 自学习\ndocs/spec/sut-self-learning.md"]
+  Q -->|"学习被测系统 / 探索式测试"| Learn["SUT 自学习编排\n.cursor/agents/sut-self-learning.md"]
   Q -->|"业务 UI/API 流自动化"| BDD["BDD 资产流水线（下半程）\n.cursor/agents/bdd-asset-pipeline.md"]
   Q -->|"明确要求 pytest / 性能 / DB 核对"| Pytest["pytest 路径\n复用 tuner_testkit + packages，不产出 .feature"]
   Q -->|"需要一个新的独立工具"| App["create-app skill\napps-authoring / apps-handover 规则"]
@@ -52,7 +52,7 @@ flowchart TD
   Q -->|"读/策展/索引知识"| Idx["查 INDEX.md + maintain-index skill"]
 ```
 
-- **学习被测系统** → `docs/spec/sut-self-learning.md`（explore-first：能力清单 + design 知识；证据来源开放、design 产物收敛）。
+- **学习被测系统** → `.cursor/agents/sut-self-learning.md`（编排 explore → freeze → design；规格 `docs/spec/sut-self-learning.md`；任务 log 见 `docs/spec/agent-task-log.md`）。
 - **业务 UI/API 流** → `.cursor/agents/bdd-asset-pipeline.md`（自学习的**下半程**；内部含三门禁，仅在此分支生效）。
 - **明确 pytest** → 直接写 pytest，复用 `tuner_testkit` 与本仓 `packages/**` 业务资产，不要为满足分层而硬造 Gherkin。
 - **造工具** → `create-app` skill；工具做好按 `apps-handover.mdc` 补交接文档。证据通道（日志/DB 日志取数）也落 `apps/`，遵守 CLI + JSON 输出约定。
@@ -65,11 +65,14 @@ flowchart TD
 - 不得将 token / cookie / Authorization / 密码 / session 等敏感信息写入仓库（含 capture 与 api_objects）。
 - 提交遵循 Conventional Commits + 六层 scope，见 `docs/spec/commit-convention.md`（`commit-msg` hook 会校验）。
 - DB 只走 `tuner_testkit.db`，日志只走 `tuner_testkit.logging`（见对应 rule）。
-- 密钥只放本机 `config/env_local.py` 或环境变量，**任何分支都不得提交**。多套 prd/uat/dev 写在 `env_local.ENVIRONMENTS` 里，用 `python -m tuner_testkit.config use <name>` 切换（或 `TUNER_ENV`）。
+- 密钥只放本机 `config/env_local.py`、`data/**/*.local.yaml` 或环境变量，**任何分支都不得提交**。多套 prd/uat/dev 写在 `env_local.ENVIRONMENTS` 里，用 `python -m tuner_testkit.config use <name>` 切换（或 `TUNER_ENV`）。探索账号优先 `data/sut-accounts.local.yaml`（模板 `data/sut-accounts.example.yaml`）。
 
 ## 6. 分支与下游仓
 
 本公开仓只维护**平台**：`tuner_testkit/`、`.cursor/**`、`docs/spec/**`、`INDEX.md`。
+**本仓不是 workspace**：真正的个人 workspace 是 `tuner-init scaffold` 生成的下游仓。仓内 `dogfood/`
+是用同一 scaffold 生成的 workspace 示范（被测系统 = 平台本身），存放路线图、验收剧本与活文档；
+它是 uv workspace 成员，不带自己的 `.cursor/`（根 DNA 覆盖全树），提交 scope 用 `dogfood`。
 
 | 分支 | 职责 | 谁改 |
 | --- | --- | --- |
@@ -78,3 +81,7 @@ flowchart TD
 - **下游 SUT 测试仓**用 `tuner-init scaffold` 从空白目录新建，用 `uv add tuner-testkit==x.y.z` + `tuner-dna sync` 取平台能力。业务资产（`assets/`、objects、features、`INDEX.project.md`）只活在下游仓。
 - 试用中发现框架要改：在 `public-main` 改并提交（内循环可用 `uv tool install --from <this-repo>`）；阶段收口打 tag 发 PyPI，下游再升级。
 - 历史分支 `plane-dogfood` **退役**：分支式 dogfood 验不了 `tuner-init` + PyPI 全链。Plane 的业务测试仓是外部仓（如 `tuner-test-platform-testrepo`）。
+- **「Plane」在本仓只指 dogfood 的被测系统**（plane.so 类项目管理产品）。曾经的 Plane Job / Sync 协议
+  （`@plane_app`、`@plane_*`、`index_platform`）已在 5.0.0 退场，对应能力中立化为 `tuner_testkit.tools`
+  （tool manifest）与 `tuner_testkit.catalog`（`tuner-workspace catalog`）；云端拉取 workspace 属于 D 阶段，
+  见 `dogfood/assets/domain-notes/platform/roadmap.md`。