Lark FE Skills

English | 中文

Lark/Feishu based Codex skills for frontend requirement planning.

Install · Prerequisites · CLI · Roadmap · 中文

English

✨ What is this?

@renkosky/lark-fe-skills packages Codex skills for Lark-based frontend workflows.

The first packaged workflow includes prd-fe-task for task breakdowns and prd-fe-schedule for Lark Base schedule dry-runs.

🚀 Install

1. Install the package:

npm install -g @renkosky/lark-fe-skills

2. Install the Codex skills:

lark-fe-skills install

3. Restart Codex or open a new Codex session.

PATH fallback:

npm exec --package=@renkosky/lark-fe-skills -- lark-fe-skills install

Use it in Codex:

[$prd-fe-task](~/.codex/skills/prd-fe-task/SKILL.md) +config
[$prd-fe-task](~/.codex/skills/prd-fe-task/SKILL.md) +plan <Lark document title, URL, or token>

🧰 Prerequisites

Install Node.js and configure lark-cli first:

npm --version
npm install -g @larksuite/cli
lark-cli config init
lark-cli auth login --recommend

Official lark-cli documentation: https://github.com/larksuite/cli/tree/main

🖥️ CLI

lark-fe-skills list
lark-fe-skills path
lark-fe-skills path <skill-name>
lark-fe-skills install
lark-fe-skills install <skill-name>
lark-fe-skills uninstall
lark-fe-skills uninstall <skill-name>
lark-fe-skills uninstall lark-fe-task

📦 Packaged Skills

| Skill | Description | | --- | --- | | prd-fe-task | Turn Lark/Feishu requirement documents into concrete frontend task breakdowns. | | prd-fe-schedule | Preview Lark Base schedule records from generated frontend task Markdown. |

📝 Notes

• npm installation and Codex skill installation are separate steps.
• This package installs skills to ~/.codex/skills.
• Use lark-fe-skills uninstall to reset local Codex skill installation for first-run testing.
• Use lark-fe-skills uninstall lark-fe-task to remove the legacy install directory.
• Existing Lark skills may live under ~/.agents/skills; that is a different source.

🗺️ Roadmap

• [x] Phase 1: Generate frontend task breakdown Markdown from Lark requirement documents.
  • [x] Configure repository root, frontend project paths, and output directory.
  • [x] List configured repositories from the global registry.
  • [x] Fetch Lark requirement documents by title, URL, or token.
  • [x] Inspect configured frontend project paths lightly for code-location clues.
  • [x] Generate ignored Markdown task files under the configured output directory.
• [ ] Phase 2: Create Lark schedule tasks from generated frontend task Markdown.
  • [x] Add prd-fe-schedule dry-run skill.
  • [x] Add YAML schedule profiles with comments for configurable fields, defaults, and record rules.
  • [x] Accept a Lark schedule Base URL or title, then read its table schema before generating a profile.
  • [x] Map Markdown tasks into schedule fields using profile rules instead of hard-coded company task types.
  • [x] Configure and remember task classification rules with +config-types.
  • [x] Read custom task status/type options from the Base instead of inventing tags.
  • [x] Support dry-run previews from a profile without modifying Lark.
  • [x] Reduce token usage by printing compact dry-run summaries and writing full previews to local temp files.
  • [x] Add an explicit +create helper that requires confirmation/--yes before writing records.
• [ ] Phase 3: Implement frontend tasks from generated Markdown.
  • [ ] Default to implementing one task at a time to keep human review small and manageable.
  • [ ] After each task, stop and report changed files, verification results, remaining tasks, mocks used, and unresolved questions.
  • [ ] Continue to the next task only when the user explicitly asks to continue, or when the user explicitly requests multiple tasks in one batch.
  • [ ] If one task is too large, split it into smaller implementation batches before editing code.
  • [ ] Reuse existing project APIs, hooks, stores, components, permissions, i18n, and mock patterns before adding new code.
  • [ ] Do not perform real backend integration in this phase; use existing reusable data or the project's mock system when API behavior is unavailable.
  • [ ] Ask the user when required API fields, permission keys, status enums, or core product behavior cannot be found in the codebase.
  • [ ] Keep changes scoped to the affected project paths and modules from the task Markdown.
  • [ ] Verify each completed task with the most relevant available static checks, and clearly report any unrelated existing failures.
• [ ] Phase 4: Backend integration and API handoff workflow.
  • [ ] Treat this phase as TBD until the backend handoff format is standardized.
  • [ ] Expected direction: use Lark as the API handoff and collaboration surface.
  • [ ] Before implementation, align with backend engineers on the required document format, including endpoint, method, request params, response schema, error codes, mock examples, and owner.
  • [ ] After the format is confirmed, extend this skill to read the API handoff document and update frontend tasks or implementation notes accordingly.

中文

安装 · 前置条件 · CLI · 开发规划 · English

✨ 这是什么？

@renkosky/lark-fe-skills 用于分发基于 Lark/飞书的前端需求工作流 Codex skills。

首个工作流包含 prd-fe-task 和 prd-fe-schedule：前者生成前端任务拆分，后者基于任务 Markdown 生成 Lark Base 排期 dry-run 预览。

🚀 安装

1. 安装 npm 包：

npm install -g @renkosky/lark-fe-skills

2. 安装 Codex skills：

lark-fe-skills install

3. 重启 Codex 或新开一个 Codex 会话。

PATH fallback：

npm exec --package=@renkosky/lark-fe-skills -- lark-fe-skills install

在 Codex 中使用：

[$prd-fe-task](~/.codex/skills/prd-fe-task/SKILL.md) +config
[$prd-fe-task](~/.codex/skills/prd-fe-task/SKILL.md) +plan <飞书文档标题、链接或 token>

🧰 前置条件

先准备 Node.js，并安装配置 lark-cli：

npm --version
npm install -g @larksuite/cli
lark-cli config init
lark-cli auth login --recommend

官方 lark-cli 文档：https://github.com/larksuite/cli/tree/main

🖥️ CLI

lark-fe-skills list
lark-fe-skills path
lark-fe-skills path <skill-name>
lark-fe-skills install
lark-fe-skills install <skill-name>
lark-fe-skills uninstall
lark-fe-skills uninstall <skill-name>
lark-fe-skills uninstall lark-fe-task

📦 已包含 Skills

| Skill | 说明 | | --- | --- | | prd-fe-task | 将 Lark/飞书需求文档拆解成可开发的前端任务。 | | prd-fe-schedule | 根据前端任务 Markdown 预览 Lark Base 排期记录。 |

📝 说明

• npm 包安装和 Codex skill 安装是两步。
• 本包默认安装到 ~/.codex/skills。
• 可以用 lark-fe-skills uninstall 清理本地 Codex skills，方便测试首次安装流程。
• 可以用 lark-fe-skills uninstall lark-fe-task 清理旧版安装目录。
• 已有 Lark skills 可能位于 ~/.agents/skills，这是另一类来源。

🗺️ 开发规划

• [x] Phase 1：根据 Lark 需求文档生成前端任务拆分 Markdown。
  • [x] 配置仓库根目录、前端项目路径和输出目录。
  • [x] 从全局 registry 查看已配置仓库。
  • [x] 支持通过标题、链接或 token 读取 Lark 需求文档。
  • [x] 轻量检查已配置前端项目路径，补充代码定位线索。
  • [x] 在已配置输出目录下生成被 git 忽略的 Markdown 任务文件。
• [ ] Phase 2：把已生成的前端任务 Markdown 转成 Lark 排期任务。
  • [x] 新增 prd-fe-schedule dry-run skill。
  • [x] 新增带注释的 YAML schedule profile，用于配置字段、默认值和记录生成规则。
  • [x] 支持用户提供 Lark 排期 Base 链接或名称，并在生成 profile 前读取表结构。
  • [x] 根据 profile 规则映射 Markdown 任务，不再写死某家公司任务类型。
  • [x] 支持通过 +config-types 配置并记住任务分类规则。
  • [x] 从 Base 读取自定义任务状态/类型选项，不臆造 tag。
  • [x] 支持基于 profile 的 dry-run 预览，不修改 Lark。
  • [x] 默认输出精简 dry-run 摘要，并把完整预览写入本地临时文件，降低 token 消耗。
  • [x] 新增显式 +create helper，要求确认/--yes 后才写入记录。
• [ ] Phase 3：根据已生成的 Markdown 实际开发前端任务。
  • [ ] 默认一次只实现一个 Task，控制代码改动粒度，方便人工 review。
  • [ ] 每完成一个 Task 就暂停，并汇报改动文件、验证结果、剩余任务、使用的 mock 和未解决问题。
  • [ ] 只有用户明确要求继续，或明确要求一次实现多个任务时，才继续处理后续 Task。
  • [ ] 如果单个 Task 本身过大，先拆成更小的实现批次再改代码。
  • [ ] 优先复用项目里已有的接口、hooks、stores、组件、权限、i18n 和 mock 模式，再考虑新增实现。
  • [ ] 这一阶段不做真实后端接口对接；接口行为不可用时，优先使用已有可复用数据或项目 mock 体系。
  • [ ] 如果 API 字段、权限 key、状态枚举或核心产品行为在代码里找不到，要先询问用户，不臆造。
  • [ ] 代码改动只限制在任务 Markdown 标出的受影响项目路径和模块内。
  • [ ] 每完成一个 Task 后运行最相关的静态检查，并明确说明是否存在无关历史失败。
• [ ] Phase 4：接口对接与后端交付文档工作流。
  • [ ] 这部分在后端接口交付格式统一前保持待定。
  • [ ] 预期方向：继续使用 Lark 作为接口交付和协作载体。
  • [ ] 实施前需要先和后端约定文档格式，包括接口地址、请求方法、请求参数、响应结构、错误码、mock 示例和负责人。
  • [ ] 格式确认后，再扩展 skill 读取接口交付文档，并同步更新前端任务或开发说明。