lark-remote 0.2.0 → 0.2.1

package/CHANGELOG.md

@@ -5,6 +5,16 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.2.1] - 2026-09-11
+
+### 新增
+
+- **Windows 正式支持**：原生 Windows 支持已在真机完成验证，README / getting-started 的平台前置条件从「macOS / Linux，暂不支持 Windows」更新为 macOS / Linux / Windows 10+
+
+### 变更
+
+- **安全文档措辞更新**：README 安全警告与 `SECURITY.md` 反映 Claude 审批模式——`permissionMode` 可配置（默认 `bypassPermissions`，配置为其他值即启用飞书卡片交互式审批），不再表述为「固定 bypassPermissions」；审批卡片描述从 Codex 专属改为 Claude / Codex / Kimi 通用
+
 ## [0.2.0] - 2026-09-11
 
 ### 新增
@@ -62,7 +72,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### 变更
 
 - kimi ACP 能力声明：`fs` 保持关闭（走本地磁盘兜底，避免未实现的 `fs/*` RPC 报 `METHOD_NOT_FOUND`），`terminal` 打开
-- 文档清理：移除设计文档中对内部文档名（`lessons-redlines`）与本地仓库路径（`/Users/.../lark-remote`）的引用
 
 ## [0.1.10] - 2026-08-22
 
@@ -212,7 +221,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - `displayTitle`/`summary` 跳过 task-notification 注入
 - 修复 `scanClaudeSessions` 对迁移到其他目录的会话的遗漏（S2）
 - 修复 probe 生命周期、工厂回退、卡片回调等评审发现的问题（F1-F8）
-- 清理引用已删除文档（lessons-redlines、resume-card-budget-failure-analysis）导致的失效链接
 
 ### 变更
 
@@ -221,7 +229,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - 移除 `pnpm-lock.yaml`，统一使用 `bun.lock` 管理依赖
 - 新增 husky lint-staged pre-commit / pre-push 钩子（提交前 lint + typecheck）
 - 启用 npm provenance（构建产物可溯源）
-- 移除 watchdog 相关用户文档及 `open-source-playbook` 等临时文档
 
 ### 测试
 
@@ -300,3 +307,4 @@ Initial release.
 [0.1.6]: https://github.com/bungabungawoda/lark-remote/releases/tag/v0.1.6
 [0.1.7]: https://github.com/bungabungawoda/lark-remote/releases/tag/v0.1.7
 [0.1.8]: https://github.com/bungabungawoda/lark-remote/releases/tag/v0.1.8
+[0.2.1]: https://github.com/bungabungawoda/lark-remote/releases/tag/v0.2.1

package/README.en.md

@@ -16,20 +16,20 @@ Designed for a single-user, peer-to-peer private chat. The agent's system prompt
 
 ## ⚠️ Security Warning (read first)
 
-This tool turns Feishu messages directly into local agent executions. Claude runs with **full permissions** by default (fixed `bypassPermissions`); Codex uses the **app-server approval mode** (default `approvalPolicy=on-request` — command execution requires your confirmation on the Feishu card, sandbox defaults to `workspace-write`), which is relatively safe; other agents run with their CLI defaults. Agents can still read and write the local directories you specify. You MUST:
+This tool turns Feishu messages directly into local agent executions. **Out of the box, agents run without approval prompts**: Claude defaults to `bypassPermissions` (no approval card), but that permission mode is **configurable** — set it to anything other than `bypassPermissions` (`default` / `acceptEdits` / `auto` / `dontAsk` / `plan`) to enable interactive approval, where high-risk actions wait for your confirmation on the Feishu card. Codex uses the **app-server approval mode** (default `approvalPolicy=on-request` — command execution requires your confirmation on the Feishu card, sandbox defaults to `workspace-write`); Kimi defaults to `manual` (per-action approval); other agents run with their CLI defaults. **In any mode**, agents can read and write the local directories you specify. You MUST:
 
 - **Use it only in your own private (p2p) chat.** Never add the bot to any group chat; never let anyone else talk to it.
 - Restrict the app's visibility to **yourself only** in the Feishu Open Platform.
 - Don't run it on a machine holding data you can't afford to lose; prefer a dedicated user / machine / container.
 - Keep credentials such as `appSecret` in your local `config.yaml` only — **never commit them anywhere**.
 
-In Codex approval mode, an approval card arrives before any command runs — approve or reject it right in Feishu:
+In approval mode (Claude, Codex, Kimi), an approval card arrives before a command runs — approve or reject it right in Feishu:
 
-![Codex command approval card example](docs/images/approval-card.png)
+![Command approval card example](docs/images/approval-card.png)
 
 ## Prerequisites
 
-- OS: macOS / Linux. **Windows is not supported yet** (relies on POSIX behaviors such as Unix signals, bash, and file locks)
+- OS: macOS / Linux / Windows 10+ (native Windows support is verified on real hardware, with behavior matching macOS/Linux)
 - Node.js 20+ ([Bun](https://bun.sh/) for development)
 - A Feishu custom app (either way): scan-to-create (a QR code pops up in the terminal on first launch; scan it with the Feishu app to create the app and write credentials automatically); or create one manually in the Open Platform — enable bot capability, subscribe to `im.message.receive_v1` and `card.action.trigger` via **long connection** (WebSocket, no public address needed), with at least the `im:message` scope.
 - Claude Code CLI installed locally and logged in once in a terminal (`claude` → browser OAuth). Same idea for other agents (codex / opencode / pi / kimi): install the corresponding CLI first. DSH connects to a local DSH Web Host via HTTP+WebSocket (no local subprocess), default address `http://127.0.0.1:3080`.
@@ -70,7 +70,7 @@ defaultAgent: claude
 claude:
   model: claude-opus-4-8
   effort: medium            # low | medium | high | xhigh | max
-  # permissionMode is hardcoded to bypassPermissions (inside the runner)
+  permissionMode: bypassPermissions  # Claude official --permission-mode: default | acceptEdits | auto | bypassPermissions | manual | dontAsk | plan (switchable via /config card; anything other than bypassPermissions enables approval)
   stopGraceMs: 5000
 
 logging:

package/README.md

@@ -16,20 +16,20 @@
 
 ## ⚠️ 安全警告（必读）
 
-本工具会把飞书消息直接转为本地 agent 执行。Claude 默认以**完全权限**运行（固定 `bypassPermissions`）；Codex 采用 **app-server 审批模式**（默认 `approvalPolicy=on-request`，执行命令需在飞书卡片上确认，沙箱默认 `workspace-write`），相对安全；其他 agent 以各自 CLI 的默认权限运行。agent 仍可读写你指定的本地目录，请务必：
+本工具会把飞书消息直接转为本地 agent 执行。**开箱默认是不弹审批的**：Claude 默认以 `bypassPermissions` 运行（无审批卡），但该权限模式**可配置**——把它改成 `bypassPermissions` 以外的值（`default` / `acceptEdits` / `auto` / `dontAsk` / `plan`）即启用交互式审批，高风险操作会在飞书卡片上等你确认；Codex 采用 **app-server 审批模式**（默认 `approvalPolicy=on-request`，执行命令需在飞书卡片上确认，沙箱默认 `workspace-write`）；Kimi 默认 `manual`（逐条审批）；其他 agent 以各自 CLI 的默认权限运行。**无论哪种模式**，agent 都能读写你指定的本地目录，请务必：
 
 - **仅限你自己的私聊（p2p）使用**。不要把机器人拉进任何群聊，不要让任何其他人能与它对话。
 - 在飞书开放平台把应用可见范围限制到**只有你自己**。
 - 运行本工具的机器上不要存放与损失承受力不符的数据；建议在独立用户/机器/容器里运行。
 - `appSecret` 等凭据只在本地 `config.yaml`，**不要提交到任何仓库**。
 
-Codex 审批模式下，执行命令前会收到审批卡片，可直接在飞书里允许或拒绝：
+审批模式下（Claude、Codex、Kimi），执行命令前会收到审批卡片，可直接在飞书里允许或拒绝：
 
-![Codex 命令审批卡片示例](docs/images/approval-card.png)
+![命令审批卡片示例](docs/images/approval-card.png)
 
 ## 前置条件
 
-- 操作系统：macOS / Linux。**暂不支持 Windows**（原生 Windows 支持在建设中：平台 seam 已落地，待真机验证后发布）
+- 操作系统：macOS / Linux / Windows 10+（原生 Windows 支持已完成真机验证，行为与 macOS/Linux 对齐）
 - Node.js 20+（开发时用 [Bun](https://bun.sh/)）
 - 飞书自建应用（二选一）：扫码创建（首次启动终端弹二维码，飞书 App 扫码即自动创建并写入凭据）；或手动在开放平台建应用——开启机器人能力，订阅 `im.message.receive_v1` 和 `card.action.trigger`，订阅方式选「长连接」（WebSocket，无需公网地址），权限至少 `im:message`。
 - Claude Code CLI：本地安装并在终端完成一次登录（`claude` → 浏览器 OAuth）。使用其他 agent（codex / opencode / pi / kimi）同理，先装好对应 CLI；DSH 需本地 DSH Web Host 在跑（默认 `http://127.0.0.1:3080`），lark-remote 通过 HTTP+WebSocket 直连、不 spawn 本地子进程。
@@ -70,7 +70,7 @@ defaultAgent: claude
 claude:
   model: claude-opus-4-8
   effort: medium            # low | medium | high | xhigh | max
-  # permissionMode 硬编码为 bypassPermissions（runner 内部）
+  permissionMode: bypassPermissions  # Claude 官方 --permission-mode：default | acceptEdits | auto | bypassPermissions | manual | dontAsk | plan（`/config` 卡片可切换；非 bypassPermissions 会启用审批）
   stopGraceMs: 5000
 
 logging:

package/dist/bridge/index.js

@@ -699,7 +699,7 @@ export class Bridge {
             const workspaces = this.workspaceStore.list();
             if (workspaces.length > 0) {
                 // NOTE: fallback uses insertion order, not sort preference — by design
-                // (workspace-sorting.md §9: cwd fallback stays insertion-order for now)
+                // (cwd fallback stays insertion-order for now)
                 cwd = workspaces[0].path;
                 getLogger().info(`[bridge] resolveCwd fallback cwd=${cwd} (first saved workspace)`);
             }

package/dist/connector/index.js

@@ -130,7 +130,6 @@ export class FeishuConnector {
         // 返回时，lark SDK 正常 resolve、@larksuite/channel 的 patchCard 丢弃返回值，
         // 导致终态卡 patch 被业务层拒绝时全链路无日志无兜底。这里只观测不改行为——
         // 不 throw、不重试，返回值原样透传。
-        // 详见 .adversarial-tdd/prompt-patchcard-business-code-observability.md
         // Guard: unit-test mocks may omit rawClient; skip probe installation in that case.
         const messageService = tryGetPatchService(this.channel);
         if (messageService?.patch) {

package/dist/index.js

@@ -438,7 +438,7 @@ function setupMessageHandlers(connector, router, bridge, sessionStore, workspace
             const workspaces = workspaceStore.list();
             if (workspaces.length > 0) {
                 // NOTE: fallback uses insertion order, not sort preference — by design
-                // (workspace-sorting.md §9: cwd fallback stays insertion-order for now)
+                // (cwd fallback stays insertion-order for now)
                 workspace = workspaces[0].path;
             }
         }

package/dist/router/index.js

@@ -1029,7 +1029,7 @@ export class CommandRouter {
         }
         const resolvedTarget = path.resolve(targetPath);
         // 与 ls.browse 对齐：目标只要存在且是目录即可切换（不再做子树/父级限制）。
-        // 安全边界由 binder owner 认证兜底，见 docs/architecture/design.md §9.7 更新说明。
+        // 安全边界由 binder owner 认证兜底，见 docs/zh/architecture/design.md §9.7 更新说明。
         if (!fs.existsSync(resolvedTarget) || !fs.statSync(resolvedTarget).isDirectory()) {
             await this.bridge.sendResult({ text: `路径无效: ${resolvedTarget}` }, ctx);
             return;

package/dist/session/claude/sessions.js

@@ -114,7 +114,7 @@ export function listClaudeSessions(cwd, opts = {}) {
     const limit = opts.limit;
     const projectsDir = opts.projectsDir ?? defaultProjectsDir();
     const index = ensureSessionIndex(projectsDir);
-    // 首次调用延迟全量建索引；之后 5s 按需节流增量刷新（session-index-manual 拍板 3）。
+    // 首次调用延迟全量建索引；之后 5s 按需节流增量刷新（拍板 3）。
     index.refresh();
     // 统一 parser 的完整 cwd 集合 → A→B→C 任意中间 cwd 都可列出（核心 bug 修复）。
     // readdir 失败时索引保留旧快照（首次失败则为空），与旧扫描的 readdir 兜底等价。

package/dist/session/dsh/sessions.js

@@ -24,7 +24,7 @@ import { truncateUtf8, TOOL_RESULT_MAX_BYTES } from '../../common/truncate.js';
  * CC-05: history 拉取上限。DSH session.history 的 maxMessages 同时限制返回条数，
  * 若用它同时充当展示截断，>50 事件的会话累计 usage 会少算。这里拉取一个足够大的
  * 上限保证 usage 累加覆盖整个会话；展示仍由 capEvents(opts?.maxEvents ?? 50) 截断。
- * （若 DSH 支持分页 cursor 可改成分页，当前用有界大拉取，见 review-fix-plan-cc.md）
+ * （若 DSH 支持分页 cursor 可改成分页，当前用有界大拉取）
  */
 const DSH_HISTORY_FETCH_LIMIT = 10_000;
 /** Synchronous unary POST via curl (the reader interface is synchronous). */

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "lark-remote",
-  "version": "0.2.0",
+  "version": "0.2.1",
   "description": "Feishu (Lark) private-chat bridge to local coding agents — Claude Code / Codex / opencode / pi / kimi, with live-streaming CardKit 2.0 cards",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",