agentic-flow-kit 2.0.0 → 2.1.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 dc
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -3,33 +3,38 @@
 AI-Native 工作流驱动开发引擎的独立安装器。把现行仓的 `.agents/` 引擎、git 门禁、宿主适配层与 `wiki`/`workflow` 数据资产脚手架，从零生成到任意项目；功能与现行引擎保持 parity。
 
 - 零运行时依赖：只用 Node.js ≥ 22 内置模块。
-- 指定 agent：一次可装一个或多个宿主壳（命令 / 角色 / 规则 / hooks 接线）。
+- 指定 agent：一次可装一个或多个宿主壳（命令 / 角色 / 规则 / hooks / 技能接线），按各宿主原生规范落盘。
+- 技能单源：技能正文只保留在 `.agents/skills/`；原生支持技能的宿主生成指针壳，其余宿主生成 `wf-skills` 索引。
+- 模型按角色配置：安装时探测宿主配置 / 环境变量（可选只读 CLI），按 `code` / `judgment` / `review` 推荐，也可 `--model` 或 TUI 手填。
 - 可选语言：`--lang zh-CN`（默认）或 `--lang en`；语言包覆盖文档、提示词、记忆文件、脚手架与安装器输出。
 - 幂等升级：`.agents/install-manifest.json` 记录版本、宿主、语言与文件哈希；冲突默认跳过，`--force` 才覆盖。
+- 交互式 TUI 向导：交互终端不带参数直接运行进入菜单（安装 / 升级 / 追加宿主 / 体检 / 宿主列表），零依赖；非交互终端自动降级为编号问答或打印帮助。
 
 ## 快速开始
 
-kit 源码独立于业务仓，位于 `E:/Demo/cli-tools/agentic-flow-kit`。在目标项目根运行：
+在目标项目根运行：
 
 ```bash
-node E:/Demo/cli-tools/agentic-flow-kit/bin/flow-kit.mjs install \
+npx agentic-flow-kit install \
   --agent trae --agent codebuddy \
   --lang zh-CN \
   --stack dotnet \
   --target .
 ```
 
-发布到包源后也可以：
+从本仓库源码运行时，把 `npx agentic-flow-kit` 换成 `node <仓库路径>/bin/flow-kit.mjs`。
+
+也可以用交互式向导：在交互终端不带任何参数直接运行，进入 TUI 菜单，按提示选宿主 / 语言 / 栈与开关：
 
 ```bash
-npx agentic-flow-kit install --agent claude --lang en --target .
+npx agentic-flow-kit
 ```
 
 安装后：
 
 ```bash
 node .agents/scripts/check-local-gates.mjs --fix
-node E:/Demo/cli-tools/agentic-flow-kit/bin/flow-kit.mjs doctor --target .
+npx agentic-flow-kit doctor --target .
 ```
 
 ## 子命令
@@ -48,40 +53,100 @@ node E:/Demo/cli-tools/agentic-flow-kit/bin/flow-kit.mjs doctor --target .
 |---|---|
 | `--target <dir>` | 目标项目根，默认当前目录 |
 | `--lang zh-CN\|en` | 语言包，默认 `zh-CN` |
-| `--stack dotnet\|node\|empty` | 项目栈 profile；缺省按 `.sln`/`.csproj`/`package.json` 探测 |
+| `--stack dotnet\|node\|typescript\|rust\|go\|java\|empty` | 项目栈 profile；缺省按根目录标记探测：`.sln`/`.csproj` → dotnet，`Cargo.toml` → rust，`go.mod` → go，`pom.xml`/`build.gradle` → java，`tsconfig.json` → typescript，`package.json` → node |
+| `--tier lean\|standard\|enterprise` | 安装档位。`lean` 不生成 CI，适合小项目；`standard`（默认）生成 CI；`enterprise` 另启用加固配置（sandbox / 凭据读拒绝 / MCP 白名单 / 版本基线） |
+| `--hardened` | 等价于 `--tier enterprise`；加固配置写入 `.claude/settings.json`，其中 managed-only 键需由平台/IT 以 managed settings 部署才不可覆盖 |
 | `--name <name>` | 渲染 `{{projectName}}`，默认目标目录名 |
 | `--dry-run` | 只打印计划，不写盘 |
 | `--force` | 覆盖冲突/非本工具安装的既有文件 |
 | `--no-git-config` | 不自动写三条例行 git 配置 |
 | `--track-hosts` | 不把本机宿主目录写进 `.gitignore` |
+| `--model [host:]role=<id>` | 可重复。指定模型角色；`role` 为 `code` / `judgment` / `review`，省略宿主则全局生效 |
+| `--model-mode recommend\|detect\|inherit` | `recommend`（默认，探测 + 推荐）、`detect`（强制重新探测）、`inherit`（全部继承宿主默认） |
+| `--probe-models` | 允许执行宿主注册表中声明的只读模型列表命令（1.5s 超时，失败静默） |
+| `--no-model-config` | 跳过模型探测 / 注入，也不写 `.agents/config/model-roles.local.json` |
 
 ## 宿主
 
 v1 支持 10 个现行宿主：
 
-| 宿主 | 目录 | 说明 |
-|---|---|---|
-| `trae` | `.trae/` | 命令 / 角色 / 规则 / hooks.json |
-| `codebuddy` | `.codebuddy/` | 命令 / 角色 / 规则 / 技能 / settings / 记忆 |
-| `claude` | `.claude/` + 根 `CLAUDE.md` | 命令 / 角色 / settings / 根记忆 |
-| `opencode` | `.opencode/` | 命令 / 角色 |
-| `qwen` | `.qwen/` | 命令 / 角色 / settings |
-| `pi` | `.pi/` | 角色 / 扩展 / settings |
-| `cline` | `.cline/` | workflows / 角色 YAML / hook |
-| `commandcode` | `.commandcode/` | 命令 / 角色 / settings |
-| `qoder` | `.qoder/` | 命令 / 角色 / settings |
-| `zcode` | `.zcode/` | 角色 / config / workflows |
+| 宿主 | 命令 | 子代理 | 技能 | hooks | 模型注入 |
+|---|---|---|---|---|---|
+| `trae` | `.trae/commands/` | `.trae/agents/*.md` | index → `wf-skills` | `hooks.json` | 否 |
+| `codebuddy` | `.codebuddy/commands/` | `.codebuddy/agents/*.md` | native `.codebuddy/skills/<name>/SKILL.md` | `settings.json` | 否 |
+| `claude` | `.claude/commands/` | `.claude/agents/*.md`（`model`） | native `.claude/skills/<name>/SKILL.md` | `settings.json` | 是（`inherit` / 别名） |
+| `opencode` | `.opencode/commands/` | `.opencode/agents/*.md`（`model`） | native `.opencode/skills/<name>/SKILL.md` | 插件式（保持现状） | 是（省略 = 继承） |
+| `qwen` | `.qwen/commands/` | `.qwen/agents/*.md`（`model`） | shared `.agents/skills`（不复制） | `settings.json` | 是（`inherit`） |
+| `pi` | 无 | `.pi/agents/*.md`（`model`） | index → `.pi/SKILLS.md` | `extensions/` | 是（网关 ID，无 `inherit`） |
+| `cline` | `.cline/workflows/` | `.cline/agents/*.yml` | index → `wf-skills` | `hooks/` 适配器 | 否 |
+| `commandcode` | `.commandcode/commands/` | `.commandcode/agents/*.md`（`model`） | shared `.agents/skills`（不复制） | `settings.json` | 是（`inherit`） |
+| `qoder` | `.qoder/commands/` | `.qoder/agents/*.md` | index → `wf-skills` | `settings.json` | 否 |
+| `zcode` | `.zcode/workflows/*.dwf.ts` | `.zcode/agents/*.md` | index → `wf-skills` | `config.json` | 否 |
 
 宿主壳全部是指针壳：正文权威仍在 `.agents/commands/`、`.agents/roles/`、`.agents/skills/`，安装器不复制正文，避免双源漂移。
 
+技能安装分三种模式（由注册表 `skills.mode` 驱动）：
+
+- **native**：宿主会自行发现 `<宿主目录>/skills/<name>/SKILL.md`。安装器按 `.agents/skills/*/SKILL.md` 的语言覆盖后内容动态生成指针壳（原生 frontmatter + 指向权威正文），不手写 22 × N 份静态副本。
+- **shared**：宿主原生读取 `.agents/skills`（Qwen、Command Code）。不生成宿主技能副本，只生成 `wf-skills` 索引入口。
+- **index**：无法确认原生技能目录的宿主生成一个索引（有命令目录时是 `wf-skills.md`，否则是 `SKILLS.md`），逐行列出技能名、描述与 `.agents/skills/<name>/SKILL.md` 路径。
+
+## 模型配置
+
+安装器按三个角色为子代理选模型：
+
+| 角色 | 用途 | 典型分工 |
+|---|---|---|
+| `code` | 实现 / 改代码 | 低成本、高吞吐（如 `haiku` / `sonnet`） |
+| `judgment` | 判断 / 规划 / 仲裁 | 强推理（如 `opus`） |
+| `review` | 独立复核 | 强推理，且建议与 `code` 不同 |
+
+解析优先级（高 → 低）：
+
+1. CLI `--model [host:]role=<id>`
+2. TUI 交互选择
+3. 已存在的 `.agents/config/model-roles.local.json`（宿主级 > 角色级）
+4. 探测 + 内置推荐
+5. 宿主 `inherit`
+
+探测来源：
+
+- 宿主配置：如 `~/.claude/settings.json`、`~/.qwen/settings.json`（JSONC 容错）。
+- 环境变量：如 `ANTHROPIC_MODEL`、`CLAUDE_CODE_SUBAGENT_MODEL`、`QWEN_MODEL`。
+- 只读 CLI：仅注册表中声明且已核实的命令（当前为 Command Code 的 `--list-models`）；默认不执行，交互安装或 `--probe-models` 才启用，1.5s 超时，失败静默。
+- 内置推荐：只写已核实的别名（当前仅 Claude 的 `sonnet` / `haiku` / `opus`），不合成未知模型 ID。
+
+解析结果写入 `.agents/config/model-roles.local.json`（用户所有，不进 manifest、升级不覆盖）；宿主子代理的 `model` 字段在渲染与语言覆盖之后注入，manifest 哈希记录最终文本。升级时若本地配置缺失（例如被 gitignore 后换机），会从已安装子代理的 `model` 字段回收，避免把用户已配置的模型静默改回 `inherit`。仅注册表声明了 `agents.modelField` 的宿主会注入；其余宿主只在输出与本地配置中记录角色模型，并提示该宿主不支持按角色指定子代理模型。
+
+`review` 与 `code` 解析为同一模型时给出 `REVIEW_SAME_AS_CODE` 警告（不阻断安装）；`doctor` 也会对现有配置做同样提示；本地配置与已安装子代理的 `model` 字段不一致时，doctor 报 `model drift`，要求先按 `.agents/skills/model-upgrade/SKILL.md` 评测再升级。
+
+## 治理、评测与记忆
+
+安装后的引擎在静态门之外还带一层 AI-Native 治理面（全部零依赖）：
+
+| 面 | 入口 | 作用 |
+|---|---|---|
+| 行为评测 | `.agents/scripts/agent-eval.mjs` + `.agents/skills/eval/` | 技能/命令/角色/模型变更后，用固定任务集跑 A/B，凭证绑定“评测敏感面”的内容签名；CI 先 `--advisory`，稳定后转硬门 |
+| 测试不可弱化 | `.agents/hooks/pre-write-check.cjs` §2.5 + `fix-task.mjs` | 修复窗口内禁改既有测试；新建复现测试放行；`TEST-GUARD-ALLOW:` 留逃生口 |
+| 离线梦境 | `.agents/scripts/dream-run.mjs` | 归档、索引、主张依据、**经验回放**、**记忆表示**、**注入层预算**、结果指标 |
+| 经验日志 | `.agents/scripts/session-log.mjs` | 记录 failure/correction/refusal，供 dreaming 对照纠错 |
+| 记忆表示 | `.agents/scripts/memory-wiki.mjs` | 从 claims/incidents/session-logs 生成 `wiki/_memory/` 链接地图 |
+| 注入层预算 | `.agents/scripts/memory-tier.mjs` | 200 行 / 25KB 上限；只出上浮候选，不自动改 |
+| 单写者租约 | `.agents/scripts/path-lease.mjs` | 并行 agent 写同一页面前加租约，防共享页丢事实 |
+| 结果指标 | `.agents/scripts/outcome-metrics.mjs` | 首次过门率、复核采纳率、逃逸缺陷、评测通过率、返工占比 |
+| 闭环状态机 | `.agents/scripts/check-pipeline-state.mjs` | stage/stopType 闭包检查；真实端到端 run 仍需小任务验证 |
+
+`--tier standard`（默认）会生成 `azure-pipelines.yml` 与 `.github/workflows/ci.yml`，在 `verify --full` 之后跑上述 advisory 检查；`lean` 不生成 CI，`enterprise` 另装加固配置。未完成项与验收口径见 [BACKLOG.md](BACKLOG.md)。
+
 ## 模板 / Profile / State 三层
 
 | 层 | 位置 | 内容 |
 |---|---|---|
 | engine | `templates/engine/` | `.agents/` 通用引擎（commands / skills / roles / hooks / scripts / config / pipelines / tests / board / notes） |
-| profile | `templates/profiles/<stack>/` | 栈相关配置与 `AGENTS.md`：`test-impact-map.json`、`lane-surfaces.txt`、`rule-budgets.txt`、`workflow-modules.txt`、`commit-check.config.json`、`delete-guard-allowlist.txt` |
+| profile | `templates/profiles/<stack>/` | 栈相关 `AGENTS.md`。栈：`dotnet` / `node` / `typescript` / `rust` / `go` / `java` / `empty` |
 | scaffold | `templates/scaffold/` | `wiki/` 目录骨架 + INDEX + 看板；`workflow/` 状态机骨架 + 空台账 + `_TEMPLATE.md` |
 | git | `templates/git/` | `.githooks/` 五钩子、`.gitattributes` 合并块、`.gitignore` 运行态块 |
+| host-meta | `templates/host-meta/` | 宿主原生技能指针 / 技能索引模板；按能力动态生成，不参与 `templates/hosts/` 静态枚举 |
 | locale | `locales/<lang>/overrides/` | 按目标路径镜像的语言覆盖；缺省回退模板 |
 
 运行态（`cache/`、`tmp/`、`*.jsonl` 台账、`confirmations.d/`、`workflow/archive`、`.agents/tools/`）不打包，由引擎在装户内产生。
@@ -109,7 +174,7 @@ v1 支持 10 个现行宿主：
   1. `engine.json` 用 `{{installedAt}}` / `{{hostsJson}}` / `{{stack}}` / `{{language}}` 占位；
   2. `check-local-gates.mjs`：dotnet 与 `TEST_DB_CONN` 检查只在 `stack=dotnet` 时执行；
   3. `check-scripts-syntax.mjs`：缺失 `scripts/verify` 时跳过（存在但为空仍 fail-closed）；
-  4. `verify.mjs`：按 `engine.json options.stack` 过滤 dotnet/Vue 专用 product 门，node 栈追加 `npm test`；
+  4. `verify.mjs`：按 `engine.json options.stack` 过滤 dotnet/Vue 专用 product 门；`node`/`typescript` 在 `package.json` 有 `test` 脚本时追加 `npm test`，`rust` 追加 `cargo test`，`go` 追加 `go test ./...`，`java` 按 `pom.xml` 或 Gradle 追加测试门；
   5. `.githooks/pre-push`：宿主接线面扩到 10 宿主；
   6. `.agents/board/dev-tour.html`：项目名改为 `{{projectName}}` 占位；
   7. `check-ledger-invariant.mjs`：本仓历史基线哈希在新仓解析不出时按无基线处理，避免装户报警。
@@ -120,9 +185,9 @@ v1 支持 10 个现行宿主：
 ## 开发
 
 ```bash
-node tools/extract-from-project.mjs --project E:/Git/Shipyard.Material   # 重新提取模板（一次性迁移工具）
+node tools/extract-from-project.mjs --project 项目目录   # 重新提取模板（一次性迁移工具）
 node tools/make-en-host-overrides.mjs                                     # 重新生成英文宿主指针壳
-node tools/parity-report.mjs --project E:/Git/Shipyard.Material          # 对比现行引擎
+node tools/parity-report.mjs --project 项目目录          # 对比现行引擎
 node --test tests/*.test.mjs                                              # 安装器测试
 ```
 
@@ -138,3 +203,7 @@ agentic-flow-kit/
 ├── tests/
 └── tools/
 ```
+
+## 许可证
+
+[MIT](LICENSE) © dc

package/bin/flow-kit.mjs

@@ -2,6 +2,8 @@
 import { main } from '../src/cli/main.mjs';
 
 main(process.argv.slice(2)).catch((err) => {
-  console.error(err && err.stack ? err.stack : String(err));
+  // 预期内的用户错误（code 已标注）只打印消息；未预期错误保留堆栈便于定位。
+  if (err?.code) console.error(err.message);
+  else console.error(err && err.stack ? err.stack : String(err));
   process.exit(1);
 });

package/locales/en/installer.json

@@ -1,5 +1,5 @@
 {
-  "help": "agentic-flow-kit v{{version}}\n\nUsage:\n  flow-kit install --agent <host> [--agent <host>...] [--lang zh-CN|en] [--stack dotnet|node|empty] [--target <dir>] [--dry-run] [--force] [--no-git-config] [--track-hosts]\n  flow-kit add-host <host> [--target <dir>] [--dry-run] [--force]\n  flow-kit upgrade [--target <dir>] [--dry-run] [--force]\n  flow-kit doctor [--target <dir>] [--deep] [--fix]\n  flow-kit list-hosts\n\nHosts: trae, codebuddy, claude, opencode, qwen, pi, cline, commandcode, qoder, zcode\n\nExample:\n  node cli-tools/agentic-flow-kit/bin/flow-kit.mjs install --agent trae --agent codebuddy --lang en --target .\n",
+  "help": "agentic-flow-kit v{{version}}\n\nUsage:\n  flow-kit install --agent <host> [--agent <host>...] [--lang zh-CN|en] [--stack dotnet|node|typescript|rust|go|java|empty] [--tier lean|standard|enterprise] [--hardened] [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--no-model-config] [--dry-run] [--force] [--no-git-config] [--track-hosts]\n  flow-kit add-host <host> [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--no-model-config] [--dry-run] [--force]\n  flow-kit upgrade [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--dry-run] [--force]\n  flow-kit doctor [--target <dir>] [--deep] [--fix]\n  flow-kit list-hosts\n\nHosts: trae, codebuddy, claude, opencode, qwen, pi, cline, commandcode, qoder, zcode\n\nModel roles: code (implementation) / judgment (reasoning) / review (independent verification; prefer different from code)\n\nExample:\n  node cli-tools/agentic-flow-kit/bin/flow-kit.mjs install --agent claude --lang en --model claude:code=haiku --model claude:review=opus --target .\n",
   "install.start": "▶ {{command}} → {{target}}",
   "install.plan": "  hosts={{hosts}} stack={{stack}} lang={{lang}} planned={{count}} files",
   "install.dryRun": "  (dry-run: print the plan only, write nothing)",
@@ -19,7 +19,92 @@
   "doctor.fail": "doctor failed: {{failed}}/{{total}} checks",
   "error.noHosts": "No host specified. Use --agent <host>; available: {{hosts}}",
   "error.unknownHost": "Unknown host: {{name}}. Available: {{hosts}}",
+  "error.badModel": "Invalid --model: {{value}} (expected [host:]role=<id>; roles: code / judgment / review)",
   "error.noManifest": "Engine is not installed in the target; run install first",
   "error.noHostsInManifest": "No hosts recorded in manifest; run install first",
-  "listHosts.title": "Supported hosts:"
+  "error.badTier": "Invalid --tier: {{value}} (expected lean / standard / enterprise)",
+  "listHosts.title": "Supported hosts:",
+  "tui.title": "agentic-flow-kit wizard",
+  "tui.menuQuestion": "Select an action",
+  "tui.menu.install": "Install",
+  "tui.menu.upgrade": "Upgrade",
+  "tui.menu.addHost": "Add host",
+  "tui.menu.doctor": "Doctor",
+  "tui.menu.listHosts": "List hosts",
+  "tui.menu.quit": "Quit",
+  "tui.hint.moveEnter": "Up/Down to select, Enter to confirm",
+  "tui.hint.toggle": "Space to toggle, Enter to confirm",
+  "tui.hostsQuestion": "Select hosts to install",
+  "tui.langQuestion": "Select language",
+  "tui.stackQuestion": "Select stack",
+  "tui.optsQuestion": "Select install options",
+  "tui.opt.gitConfig": "Configure git gates (recommended)",
+  "tui.opt.dryRun": "dry-run: print the plan only, write nothing",
+  "tui.opt.force": "Force overwrite conflicting files",
+  "tui.detected": "detected",
+  "tui.detectedFile": "detected {{file}}",
+  "tui.stack.dotnet": ".NET",
+  "tui.stack.node": "Node.js",
+  "tui.stack.typescript": "TypeScript",
+  "tui.stack.rust": "Rust",
+  "tui.stack.go": "Go",
+  "tui.stack.java": "Java",
+  "tui.stack.empty": "Unspecified",
+  "tui.currentStack": "current",
+  "tui.upgradeStackQuestion": "Select the stack for this upgrade",
+  "tui.installed": "installed",
+  "tui.planTitle": "Install plan",
+  "tui.plan.target": "Target: {{target}}",
+  "tui.plan.hosts": "Hosts: {{hosts}}",
+  "tui.plan.lang": "Language: {{lang}}",
+  "tui.plan.stack": "Stack: {{stack}}",
+  "tui.confirmInstall": "Start install?",
+  "tui.confirmUpgrade": "Start upgrade?",
+  "tui.askDoctor": "Run doctor right after install?",
+  "tui.running": "Running…",
+  "tui.done": "Done",
+  "tui.dryDone": "Dry-run done, nothing written",
+  "tui.doctorQuestion": "Select doctor mode",
+  "tui.doctor.normal": "Standard checks",
+  "tui.doctor.deep": "Deep checks (includes check-local-gates)",
+  "tui.addHostQuestion": "Select hosts to add",
+  "tui.backOrQuit": "Back to menu?",
+  "tui.errorTitle": "Error",
+  "tui.yes": "Yes",
+  "tui.no": "No",
+  "tui.inputNumber": "Type a number and press Enter:",
+  "tui.toggleNumber": "Number toggles a choice, empty Enter finishes:",
+  "tui.invalidNumber": "Invalid number, try again",
+  "tui.minSelection": "Select at least {{min}}",
+  "install.models": "Models: {{models}}",
+  "install.modelWarning": "Model warning: {{host}} {{role}} equals the implementation model; heterogeneous review unavailable",
+  "install.localModels": "Local model config: {{path}}",
+  "install.tier": "Tier: {{tier}} (hardened: {{hardened}})",
+  "doctor.modelConfigInvalid": "model-roles.local.json cannot be parsed",
+  "doctor.reviewSameAsCode": "review equals code; heterogeneous review unavailable",
+  "doctor.tier": "Tier: {{tier}}",
+  "doctor.ciMissing": "Missing CI template: {{paths}} (ignorable for tier=lean)",
+  "doctor.guaranteeGap": "Host {{host}} does not support {{capability}}; this guarantee can only be advisory, not a deterministic gate",
+  "doctor.modelDrift": "Model drift: {{host}}/{{role}} installed {{installed}} != local {{local}}; run the model-upgrade eval before upgrading",
+  "doctor.hardenedHint": "Hardened config contains managed-only keys; deploy it as managed settings so it cannot be overridden",
+  "tui.model.useRecommended": "Use recommended models for {{host}}?",
+  "tui.model.role": "Select the {{role}} model",
+  "tui.model.role.code": "implementation (code)",
+  "tui.model.role.judgment": "judgment",
+  "tui.model.role.review": "review",
+  "tui.model.recommended": "recommended",
+  "tui.model.inherit": "inherit (follow main model)",
+  "tui.model.manual": "Enter manually…",
+  "tui.model.skip": "Skip (leave unspecified)",
+  "tui.model.manualPrompt": "Model ID:",
+  "tui.model.invalid": "Model ID cannot be empty",
+  "tui.model.allowProbe": "Run {{host}}'s read-only model list command?",
+  "tui.model.sameAsCode": "review equals code; heterogeneous review unavailable",
+  "tui.plan.models": "Models: {{models}}",
+  "tui.hint.typeEnter": "Type, then press Enter",
+  "host.skillPointer": "Authoritative skill: `{{path}}`. This file is a pointer shell; read the authoritative source at execution time. Relative paths (scripts/, references/, …) resolve from that directory.",
+  "host.skillsIndexTitle": "# Skills index",
+  "host.skillsIndexDescription": "Skills index; authoritative bodies live under .agents/skills/",
+  "host.skillsIndexLine": "- `{{name}}` — {{description}} (authoritative: `{{path}}`)",
+  "host.skillsIndexEmpty": "(no skills)"
 }

package/locales/en/overrides/.codebuddy/skills/tfs-pr/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: tfs-pr
-description: 内网自托管 TFS / Azure DevOps Server PR 技能，项目无关：自动识别 git 仓库与远端推导 collection/project/repo，PAT 缺失时引导创建并保存环境变量；查 PR / 建 PR（context 拉 diff 提炼标题与描述）/ 改 PR 描述 / 冲突预检与 worktree 隔离解决，走 REST API + PAT（az devops 扩展拒 http 组织 URL，az 不可用）。命令入口 .agents/skills/tfs-pr/scripts/tfs-pr.mjs。
+description: Internal self-hosted TFS / Azure DevOps Server PR skill, project-agnostic — auto-detects the git repository and remote to derive collection/project/repo, guides creation and saves an env var when PAT is missing; query PR / create PR (context pulls diff to distill title and description) / update PR description / conflict pre-check and worktree-isolated resolution, via REST API + PAT (the az devops extension rejects http organization URLs, so az is unavailable). Command entry .agents/skills/tfs-pr/scripts/tfs-pr.mjs.
 ---
 
 Authoritative skill: `.agents/skills/tfs-pr/SKILL.md`. This file is a pointer shell; read the authoritative source at execution time and do not copy the body here.

package/locales/en/overrides/workflow/README.md

@@ -25,6 +25,8 @@ The root `regression-checklist.md` is a **living regression checklist**: the uni
 
 **Gate judgment monthly shards** (2026-10-22 gate-metrics-shard): `workflow/gate-metrics.d/<YYYY-MM>.jsonl` holds one row per month (rule × WARN/BLOCK), generated by `node .agents/scripts/agg-judgments.mjs --emit-month` and committed with the change; pre-push monthly checks automatically backfill the previous month's shard. The retirement criterion (N consecutive months = 0) reads the union of local `.agents/judgments.jsonl` and the shards and takes the max—machine/worktree changes no longer reset it to zero; multi-machine union (`.gitattributes` reuses `jsonl-union`).
 
+**Claims ledger** (append-only, monthly): `workflow/claims.d/<YYYY-MM>.jsonl` records one sentence a future session might take as given, plus a quote that already exists in the repo. Append with `node .agents/scripts/claims.mjs add`. `recheck` only reports whether that quote is still in the file. Its exit code is always 0. It is not a gate, and it does not rewrite past lines. Merges use `jsonl-union`.
+
 ## Usage
 
 - **Start a new task**: run `.agents/commands/new-task.md` (leveling + L0 minimal path / L1 fast path entrance); L2/L3 enter the four chapters of `.agents/commands/stages.md` (Plan → Design → Build → Test, 2026-11-08 tier 2-F four-in-one)

package/locales/zh-CN/installer.json

@@ -1,5 +1,5 @@
 {
-  "help": "agentic-flow-kit v{{version}}\n\n用法：\n  flow-kit install --agent <host> [--agent <host>...] [--lang zh-CN|en] [--stack dotnet|node|empty] [--target <dir>] [--dry-run] [--force] [--no-git-config] [--track-hosts]\n  flow-kit add-host <host> [--target <dir>] [--dry-run] [--force]\n  flow-kit upgrade [--target <dir>] [--dry-run] [--force]\n  flow-kit doctor [--target <dir>] [--deep] [--fix]\n  flow-kit list-hosts\n\n宿主：trae, codebuddy, claude, opencode, qwen, pi, cline, commandcode, qoder, zcode\n\n示例：\n  node cli-tools/agentic-flow-kit/bin/flow-kit.mjs install --agent trae --agent codebuddy --lang zh-CN --target .\n",
+  "help": "agentic-flow-kit v{{version}}\n\n用法：\n  flow-kit install --agent <host> [--agent <host>...] [--lang zh-CN|en] [--stack dotnet|node|typescript|rust|go|java|empty] [--tier lean|standard|enterprise] [--hardened] [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--no-model-config] [--dry-run] [--force] [--no-git-config] [--track-hosts]\n  flow-kit add-host <host> [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--no-model-config] [--dry-run] [--force]\n  flow-kit upgrade [--target <dir>] [--model [host:]role=<id>] [--model-mode recommend|detect|inherit] [--probe-models] [--dry-run] [--force]\n  flow-kit doctor [--target <dir>] [--deep] [--fix]\n  flow-kit list-hosts\n\n宿主：trae, codebuddy, claude, opencode, qwen, pi, cline, commandcode, qoder, zcode\n\n模型角色：code（实现）/ judgment（判断）/ review（独立复核，建议与 code 不同）\n\n示例：\n  node cli-tools/agentic-flow-kit/bin/flow-kit.mjs install --agent claude --lang zh-CN --model claude:code=haiku --model claude:review=opus --target .\n",
   "install.start": "▶ {{command}} → {{target}}",
   "install.plan": "  宿主={{hosts}} 栈={{stack}} 语言={{lang}} 计划={{count}} 个文件",
   "install.dryRun": "  （dry-run：只打印计划，不写盘）",
@@ -19,7 +19,92 @@
   "doctor.fail": "doctor 失败：{{failed}}/{{total}} 项",
   "error.noHosts": "未指定宿主。用 --agent <host> 指定，可选：{{hosts}}",
   "error.unknownHost": "未知宿主：{{name}}。可选：{{hosts}}",
+  "error.badModel": "无效的 --model：{{value}}（格式：[宿主:]角色=模型，角色：code / judgment / review）",
   "error.noManifest": "目标未安装本引擎；先运行 install",
   "error.noHostsInManifest": "manifest 未登记宿主；先运行 install",
-  "listHosts.title": "支持的宿主："
+  "error.badTier": "无效的 --tier：{{value}}（可选 lean / standard / enterprise）",
+  "listHosts.title": "支持的宿主：",
+  "tui.title": "agentic-flow-kit 向导",
+  "tui.menuQuestion": "选择操作",
+  "tui.menu.install": "安装",
+  "tui.menu.upgrade": "升级",
+  "tui.menu.addHost": "追加宿主",
+  "tui.menu.doctor": "体检",
+  "tui.menu.listHosts": "宿主列表",
+  "tui.menu.quit": "退出",
+  "tui.hint.moveEnter": "↑/↓ 选择，回车确认",
+  "tui.hint.toggle": "空格勾选，回车确认",
+  "tui.hostsQuestion": "选择要安装的宿主",
+  "tui.langQuestion": "选择语言",
+  "tui.stackQuestion": "选择技术栈",
+  "tui.optsQuestion": "选择安装选项",
+  "tui.opt.gitConfig": "配置 git 门禁（推荐）",
+  "tui.opt.dryRun": "dry-run：只打印计划，不写盘",
+  "tui.opt.force": "强制覆盖冲突文件",
+  "tui.detected": "检测到",
+  "tui.detectedFile": "检测到 {{file}}",
+  "tui.stack.dotnet": ".NET",
+  "tui.stack.node": "Node.js",
+  "tui.stack.typescript": "TypeScript",
+  "tui.stack.rust": "Rust",
+  "tui.stack.go": "Go",
+  "tui.stack.java": "Java",
+  "tui.stack.empty": "未指定",
+  "tui.currentStack": "当前",
+  "tui.upgradeStackQuestion": "选择本次升级的技术栈",
+  "tui.installed": "已安装",
+  "tui.planTitle": "安装计划",
+  "tui.plan.target": "目标：{{target}}",
+  "tui.plan.hosts": "宿主：{{hosts}}",
+  "tui.plan.lang": "语言：{{lang}}",
+  "tui.plan.stack": "栈：{{stack}}",
+  "tui.confirmInstall": "开始安装？",
+  "tui.confirmUpgrade": "开始升级？",
+  "tui.askDoctor": "安装后立即体检？",
+  "tui.running": "执行中…",
+  "tui.done": "完成",
+  "tui.dryDone": "dry-run 完成，未写盘",
+  "tui.doctorQuestion": "选择体检模式",
+  "tui.doctor.normal": "普通体检",
+  "tui.doctor.deep": "深度体检（含 check-local-gates）",
+  "tui.addHostQuestion": "选择要追加的宿主",
+  "tui.backOrQuit": "返回菜单？",
+  "tui.errorTitle": "出错了",
+  "tui.yes": "是",
+  "tui.no": "否",
+  "tui.inputNumber": "输入序号后回车：",
+  "tui.toggleNumber": "输入序号切换勾选，直接回车完成：",
+  "tui.invalidNumber": "无效序号，请重试",
+  "tui.minSelection": "至少选择 {{min}} 项",
+  "install.models": "模型：{{models}}",
+  "install.modelWarning": "模型提示：{{host}} 的 {{role}} 与实现模型相同，跨模型复核不可用",
+  "install.localModels": "本地模型配置：{{path}}",
+  "install.tier": "安装档位：{{tier}}（加固：{{hardened}}）",
+  "doctor.modelConfigInvalid": "model-roles.local.json 无法解析",
+  "doctor.reviewSameAsCode": "review 与 code 相同，跨模型复核不可用",
+  "doctor.tier": "安装档位：{{tier}}",
+  "doctor.ciMissing": "缺少 CI 模板：{{paths}}（tier=lean 可忽略）",
+  "doctor.guaranteeGap": "宿主 {{host}} 不支持 {{capability}}；该保证只能是 advisory，不能当 deterministic 门",
+  "doctor.modelDrift": "模型漂移：{{host}}/{{role}} 安装值 {{installed}} ≠ 本地 {{local}}；按 model-upgrade 技能先评测再升级",
+  "doctor.hardenedHint": "加固配置含 managed-only 键；需由平台/IT 以 managed settings 部署才不可被覆盖",
+  "tui.model.useRecommended": "{{host}} 使用推荐模型？",
+  "tui.model.role": "选择 {{role}} 模型",
+  "tui.model.role.code": "实现（code）",
+  "tui.model.role.judgment": "判断（judgment）",
+  "tui.model.role.review": "复核（review）",
+  "tui.model.recommended": "推荐",
+  "tui.model.inherit": "inherit（跟随主模型）",
+  "tui.model.manual": "手动输入…",
+  "tui.model.skip": "跳过（不指定）",
+  "tui.model.manualPrompt": "输入模型 ID：",
+  "tui.model.invalid": "模型 ID 不能为空",
+  "tui.model.allowProbe": "允许执行 {{host}} 的只读模型列表命令？",
+  "tui.model.sameAsCode": "review 与 code 相同，跨模型复核不可用",
+  "tui.plan.models": "模型：{{models}}",
+  "tui.hint.typeEnter": "输入后回车",
+  "host.skillPointer": "权威技能：`{{path}}`。本文件是指针壳，执行时读取权威源；相对路径（scripts/、references/ 等）以该目录为基准。",
+  "host.skillsIndexTitle": "# 技能索引",
+  "host.skillsIndexDescription": "技能索引：权威正文在 .agents/skills/",
+  "host.skillsIndexLine": "- `{{name}}` — {{description}}（权威：`{{path}}`）",
+  "host.skillsIndexEmpty": "（暂无技能）"
 }

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "agentic-flow-kit",
-  "version": "2.0.0",
+  "version": "2.1.0",
   "description": "AI-Native workflow engine installer: generate .agents, git hooks, host adapters, wiki/workflow scaffolds from zero.",
-  "license": "UNLICENSED",
+  "license": "MIT",
   "type": "module",
   "bin": {
     "flow-kit": "bin/flow-kit.mjs"

package/schemas/install-manifest.schema.json

@@ -14,7 +14,9 @@
       "items": { "enum": ["trae", "codebuddy", "claude", "opencode", "qwen", "pi", "cline", "commandcode", "qoder", "zcode"] },
       "uniqueItems": true
     },
-    "stack": { "enum": ["dotnet", "node", "empty"] },
+    "stack": { "enum": ["dotnet", "node", "typescript", "rust", "go", "java", "empty"] },
+    "tier": { "enum": ["lean", "standard", "enterprise"] },
+    "hardened": { "type": "boolean" },
     "files": {
       "type": "object",
       "additionalProperties": { "type": "string", "pattern": "^[a-f0-9]{64}$" }

package/schemas/project-profile.schema.json

@@ -5,7 +5,7 @@
   "type": "object",
   "required": ["stack"],
   "properties": {
-    "stack": { "enum": ["dotnet", "node", "empty"] },
+    "stack": { "enum": ["dotnet", "node", "typescript", "rust", "go", "java", "empty"] },
     "ports": {
       "type": "object",
       "properties": {

package/src/cli/args.mjs

@@ -1,6 +1,6 @@
 // 极简参数解析：--key value / --flag / --key=value；--agent 可重复。
 export function parseArgs(argv) {
-  const args = { _: [], agents: [], flags: new Set() };
+  const args = { _: [], agents: [], models: [], flags: new Set() };
   const takeValue = (i) => {
     const next = argv[i + 1];
     if (next == null || (next.startsWith('--') && next !== '-')) return null;
@@ -17,6 +17,7 @@ export function parseArgs(argv) {
       const key = raw.slice(2, eq);
       const value = raw.slice(eq + 1);
       if (key === 'agent') args.agents.push(value);
+      else if (key === 'model') args.models.push(value);
       else args[key] = value;
       continue;
     }
@@ -24,6 +25,7 @@ export function parseArgs(argv) {
     const value = takeValue(i);
     if (value != null) {
       if (key === 'agent') args.agents.push(value);
+      else if (key === 'model') args.models.push(value);
       else args[key] = value;
       i += 1;
     } else {

package/src/cli/main.mjs

@@ -1,7 +1,7 @@
 import fs from 'node:fs';
 import path from 'node:path';
 import { parseArgs, flag } from './args.mjs';
-import { makeT } from '../i18n.mjs';
+import { makeT, LANGUAGES } from '../i18n.mjs';
 import { HOST_NAMES, resolveHost } from '../hosts/registry.mjs';
 import { installCommand } from '../commands/install.mjs';
 import { doctorCommand } from '../commands/doctor.mjs';
@@ -9,7 +9,7 @@ import { addHostCommand } from '../commands/add-host.mjs';
 import { upgradeCommand } from '../commands/upgrade.mjs';
 import { KIT_ROOT } from '../install/plan.mjs';
 
-const SUPPORTED_LANGS = new Set(['zh-CN', 'en']);
+const SUPPORTED_LANGS = new Set(LANGUAGES);
 
 function normalizeLang(input) {
   const lang = String(input || 'zh-CN');
@@ -27,6 +27,13 @@ async function main(argv) {
   const lang = normalizeLang(args.lang);
   const t = makeT(lang);
 
+  // 无位置参数且无任何选项（含 --agent/--target 等值形参数，它们不进 flags.size）、交互终端 → 进入 TUI 向导（懒加载，避免影响脚本路径）
+  if (!args._.length && args.flags.size === 0 && args.lang == null && args.agents.length === 0 && args.target == null
+    && process.stdin.isTTY && process.stdout.isTTY) {
+    const { wizardCommand } = await import('../tui/wizard.mjs');
+    return wizardCommand({ t, lang });
+  }
+
   if (flag(args, 'version') || flag(args, 'v')) {
     console.log(readVersion());
     return 0;