codex-to-im 1.0.10 → 1.0.12

package/README.md

@@ -10,7 +10,7 @@ The product is no longer centered around a Codex skill. The main path is:
 4. Start the bridge in the background
 5. Bind real desktop Codex threads to Feishu or Weixin chats
 
-`SKILL.md` is still kept in the repo, but only as an optional Codex integration entry.
+Optional: if you want Codex to know it can send local files or images back to IM without relying on bridge-injected prompt text, install the bundled `codex-to-im` skill from the workbench.
 
 ## Project Origin
 
@@ -19,7 +19,7 @@ The current codebase is a consolidated continuation of two earlier repositories:
 - `Claude-to-IM`
 - `Claude-to-IM-skill`
 
-`codex-to-im` is based on those two projects and has been reworked toward a single-package local app, shared-thread workflow, and optional Codex integration model.
+`codex-to-im` is based on those two projects and has been reworked toward a single-package local app and shared-thread workflow.
 
 Windows host installation guide: [docs/install-windows.md](D:/codex/Claude-to-IM-skill/docs/install-windows.md)
 
@@ -31,7 +31,6 @@ Windows host installation guide: [docs/install-windows.md](D:/codex/Claude-to-IM
 - Weixin QR login flow
 - Desktop session discovery from `~/.codex/sessions`
 - Web-side binding updates for IM chats
-- Optional Codex integration for opening `codex-to-im` or entering the Feishu handoff flow
 
 ## Install
 
@@ -190,8 +189,7 @@ This is closer to a full-power `code` workflow. It fits a controlled local proje
 The channel pages also expose a “Use Markdown for bridge feedback” switch:
 - enabled by default for Feishu
 - disabled by default for WeChat
-- only affects bridge-generated feedback such as `/h`, `/status`, and `/threads`
-- does not affect raw Codex replies
+- affects text sent through the bridge, including normal replies, shared-thread mirror messages, and system feedback such as `/h`, `/status`, and `/threads`
 
 ## Update
 
@@ -205,23 +203,6 @@ npm update -g codex-to-im
 codex-to-im
 ```
 
-## Optional Codex Integration
-
-The repo still includes a lightweight optional integration under `SKILL.md`.
-
-It is not required for the product to work.
-
-If installed into `~/.codex/skills/codex-to-im`, it should only be used for two actions:
-
-- Open `codex-to-im`
-- Open the Feishu session-sharing entry for the current workflow
-
-You can install that optional integration from the web UI, or with:
-
-```bash
-bash scripts/install-codex.sh --link
-```
-
 ## Repo Layout
 
 - `src/ui-server.ts` — local workbench UI and HTTP API
@@ -229,7 +210,6 @@ bash scripts/install-codex.sh --link
 - `src/desktop-sessions.ts` — desktop thread discovery from Codex session files
 - `src/session-bindings.ts` — binding summaries and web-side binding updates
 - `src/lib/bridge/` — bridge runtime and IM channel routing
-- `SKILL.md` — optional Codex integration only
 - `docs/` — PRD and shared-thread design docs
 
 ## Development
@@ -246,6 +226,5 @@ Current product direction:
 - Standalone local app first
 - Web workbench first
 - Shared Codex thread model first
-- Codex integration is optional, not the primary installation path
 
 [中文文档](README_CN.md)

package/README_CN.md

@@ -10,7 +10,7 @@
 4. 在后台启动 bridge
 5. 把真实的桌面 Codex thread 绑定到飞书或微信聊天
 
-仓库里仍然保留了 `SKILL.md`，但它只是一个可选的 Codex 集成入口，不再是产品本体。
+可选：如果你希望 Codex 在不依赖 bridge 注入提示词的前提下，知道“可以把本地图片/文件回发到 IM”，可以在工作台里安装附带的 `codex-to-im` skill。
 
 ## 项目来源
 
@@ -19,7 +19,7 @@
 - `Claude-to-IM`
 - `Claude-to-IM-skill`
 
-现在的 `codex-to-im` 是在这两个工程基础上继续演进的单包版本，重点调整成了本地应用、共享 thread 和可选 Codex 集成的形态。
+现在的 `codex-to-im` 是在这两个工程基础上继续演进的单包版本，重点调整成了本地应用和共享 thread 的形态。
 
 Windows 主机安装说明见：[docs/install-windows.md](D:/codex/Claude-to-IM-skill/docs/install-windows.md)
 
@@ -31,7 +31,6 @@ Windows 主机安装说明见：[docs/install-windows.md](D:/codex/Claude-to-IM-
 - 微信扫码登录
 - 从 `~/.codex/sessions` 发现桌面会话
 - 在网页中查看和切换 IM 绑定
-- 可选的 Codex 集成，仅用于打开 `codex-to-im` 或进入飞书共享入口
 
 ## 安装
 
@@ -189,8 +188,7 @@ codex-to-im stop
 通道页还支持“反馈使用 Markdown”开关：
 - 飞书默认开启
 - 微信默认关闭
-- 只影响 `/h`、`/status`、`/threads` 这类 bridge 自己生成的反馈
-- 不影响 Codex 原始回复内容
+- 影响通过 bridge 发送到通道的文本反馈，包括普通回复、共享桌面线程镜像以及 `/h`、`/status`、`/threads` 这类系统反馈
 
 ## 更新
 
@@ -204,23 +202,6 @@ npm update -g codex-to-im
 codex-to-im
 ```
 
-## 可选 Codex 集成
-
-仓库里仍然保留了一个很薄的可选集成，定义在 `SKILL.md`。
-
-它不是必需的。
-
-如果你把它装到 `~/.codex/skills/codex-to-im`，它只保留两个动作：
-
-- 打开 `codex-to-im`
-- 打开“共享当前会话到飞书”的入口
-
-你可以在 Web UI 中安装这层可选集成，也可以手动执行：
-
-```bash
-bash scripts/install-codex.sh --link
-```
-
 ## 仓库结构
 
 - `src/ui-server.ts` — 本地工作台 UI 和 HTTP API
@@ -228,7 +209,6 @@ bash scripts/install-codex.sh --link
 - `src/desktop-sessions.ts` — 从 Codex 会话文件发现桌面 thread
 - `src/session-bindings.ts` — 绑定摘要与网页侧切换
 - `src/lib/bridge/` — bridge 运行时与 IM 路由
-- `SKILL.md` — 可选 Codex 集成，不是主产品
 - `docs/` — PRD 与共享 thread 技术设计
 
 ## 开发
@@ -243,6 +223,5 @@ npm run build
 - 先做独立本地应用
 - 先做 Web 工作台
 - 先做共享 Codex thread
-- Codex 集成是可选增强，不是主安装路径
 
 [English](README.md)

package/SKILL.md

@@ -0,0 +1,67 @@
+---
+name: codex-to-im
+description: Use this skill when working through Codex-to-IM and you need Codex to send a local image or file back to the current IM chat. It teaches the attachment-send protocol and when to use it.
+---
+
+# Codex-to-IM attachment sending
+
+Use this skill only when the user is chatting through Codex-to-IM and wants you to send a generated or existing local artifact back to the IM chat.
+
+## What this skill does
+
+Codex-to-IM can send local files back to the chat after your reply. You trigger that by including one or more `<cti-send>...</cti-send>` blocks in your final answer.
+
+Supported artifact kinds:
+
+- `image`
+- `file`
+
+## Required format
+
+Output valid JSON inside the block, with no markdown fence.
+
+Single artifact:
+
+```text
+<cti-send>
+{"type":"image","path":"D:\\path\\to\\result.png","caption":"optional caption"}
+</cti-send>
+```
+
+or
+
+```text
+<cti-send>
+{"type":"file","path":"D:\\path\\to\\report.pdf","caption":"optional caption"}
+</cti-send>
+```
+
+Multiple artifacts:
+
+```text
+<cti-send>
+{"items":[
+  {"type":"image","path":"D:\\path\\to\\result.png"},
+  {"type":"file","path":"D:\\path\\to\\report.pdf"}
+]}
+</cti-send>
+```
+
+## Rules
+
+- The `path` must be an absolute local path.
+- The file must already exist before you emit the block.
+- Keep normal user-facing explanation outside the `<cti-send>` block.
+- Do not invent artifacts that were not actually created or found.
+- If you are unsure whether the file exists, verify first.
+- If the channel probably does not support the artifact well, explain that plainly instead of inventing a successful send.
+
+## When to use
+
+Use this protocol when the user asks you to:
+
+- send a generated chart, diagram, screenshot, or edited image
+- send a report, archive, patch, PDF, spreadsheet, or other output file
+- send the final result as an attachment instead of only pasting text
+
+Do not use it for ordinary text replies.

package/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "Codex to IM"
+  short_description: "Help Codex send local files back to IM chats"
+  default_prompt: "Use $codex-to-im when you need to send a generated image or file back to the current Codex-to-IM chat."

package/config.env.example

@@ -85,8 +85,8 @@ CTI_TG_CHAT_ID=your-chat-id
 # Current Codex runtime note: thinking/progress can update live, but
 # assistant body text may still arrive only at completion.
 # CTI_FEISHU_STREAMING_ENABLED=true
-# Use Markdown for bridge-generated feedback such as replies, /h, or /status.
-# Does not affect raw Codex replies.
+# Use Markdown for text sent through the bridge, including normal replies,
+# shared-thread mirror messages, and system feedback such as /h or /status.
 # CTI_FEISHU_COMMAND_MARKDOWN_ENABLED=true
 
 # ── QQ ──
@@ -111,8 +111,9 @@ CTI_TG_CHAT_ID=your-chat-id
 # only accepts WeChat-provided speech-to-text text and otherwise returns an error.
 # (default false for safety in CLI setups)
 # CTI_WEIXIN_MEDIA_ENABLED=false
-# Use Markdown for bridge-generated feedback such as replies, /h, or /status.
-# Does not affect raw Codex replies. Default is false for WeChat.
+# Use Markdown for text sent through the bridge, including normal replies,
+# shared-thread mirror messages, and system feedback such as /h or /status.
+# Default is false for WeChat.
 # CTI_WEIXIN_COMMAND_MARKDOWN_ENABLED=false
 
 # ── Permission ──