Repository navigation
fix(chat): enable disk skills in slash menu for custom ACP agents - #1121
Merged
xintaofei merged 2 commits intoNov 10, 2026
Merged
Conversation
Custom ACP agents (e.g. user-configured Antigravity instances, or custom agents with declared skill directories) do not advertise skills as ACP commands; their skills live on disk. Previously diskSkillTrigger returned null for custom agents, preventing skills from being scanned and displayed in the slash menu. Include custom agents in diskSkillTrigger so that their disk skills are scanned and listed under '/' in the composer.
…ser menus The disk-skill trigger now covers custom ACP agents, but the comments in the chat composer and the automation editor still named only Antigravity and OpenCode. Say that a custom agent's scan follows the skill directories its user declared, and that a skill it also advertises as a command is dropped by skillsForMenu.
Collaborator
|
codeg work task |
xintaofei
added a commit
that referenced
this pull request
Nov 11, 2026
This release reduces memory growth during extended use, fixes missing collaboration tools in forked Codex conversations, restores skill selection for custom agents, and updates several agents. ## Improved - **Claude Code is updated to 2.1.322 with adapter 0.106.0.** - **Updated OpenCode to 2.2.7, Kimi Code to 2.5.3, Hermes Agent to 0.23.1 and the ZCode bridge to 0.83.3.** ## Fixed - **Reduced memory buildup during repeated scrolling, view changes and streaming code generation**, improving memory usage during extended sessions. (#1114, #1115, #1117 by @kogekiplay) - **Custom agents with a configured skill directory or shared skill store now list their skills in the `/` menu**, in both chat and automation editors. (#1121 by @ana52070) - **Forked Codex conversations retain their enabled MCP tools**, allowing you to continue delegating work to sub-agents and retrieving their results. (#1105) - **Reopening a conversation no longer adds duplicate context compaction cards after the existing history.** - **OpenCode commands that time out or are terminated no longer incorrectly appear as successful.** - **OpenCode file previews correctly display reads beyond the end of a file**, while reads counted from the end no longer show a misleading line range. - **Reopening older ZCode conversations correctly restores their saved titles.** Thanks to @kogekiplay (#1114, #1115, #1117) and @ana52070 (#1121) for their contributions, and to @lxpsxx (#1105) for the report. ----------------------------- # 发布版本 0.42.3 这一版改善长时间使用时的内存占用，修复 Codex 分叉会话后协作工具丢失、自定义智能体无法选择技能等问题，并更新多个智能体。 ## 改进 - **Claude Code 更新至 2.1.322，适配器更新至 0.106.0。** - **OpenCode 更新至 2.2.7，Kimi Code 更新至 2.5.3，Hermes Agent 更新至 0.23.1，ZCode 桥接更新至 0.83.3。** ## 修复 - **减少反复滚动、切换视图和持续生成代码时的内存累积**，改善长时间使用时的内存占用。（#1114、#1115、#1117 by @kogekiplay） - **配置了技能目录或共享技能库的自定义智能体，可以在 `/` 菜单中选择技能**，聊天输入框和自动化编辑器均可使用。（#1121 by @ana52070） - **Codex 分叉会话后会保留已启用的 MCP 工具**，可以继续委派子智能体并获取执行结果。（#1105） - **重新打开会话时，已有的上下文压缩卡片不会再重复出现在历史记录末尾。** - **OpenCode 中超时或被终止的命令，不再被误标为执行成功。** - **OpenCode 读取位置超出文件末尾时，返回内容仍能正确预览**，从文件末尾倒数读取时也不会再显示错误的行号范围。 - **重新打开较早的 ZCode 会话时，能正确恢复已保存的会话标题。** 感谢 @kogekiplay（#1114、#1115、#1117）和 @ana52070（#1121）的贡献，以及 @lxpsxx（#1105）的反馈。
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
描述 (Description)
修复自定义 ACP 智能体（如自建的 Google Antigravity 实例、或配置了独立/共享技能目录的自定义 Agent）在聊天输入框中无法显示和触发磁盘技能（Disk Skills）的问题。
背景与问题根因 (Background & Root Cause)
agy_acp_server）在协议中仅上报内置命令（/plan、/logout、/usage）而从不上报技能命令，引入了diskSkillTrigger将 Antigravity 的磁盘技能挂载到输入框/菜单下。diskSkillTrigger之前仅硬编码了antigravity与open_code，对自定义智能体（isCustomAgentType(agentType)，即custom:<id>）直接返回null。agy_acp_server搭建的自定义 Antigravity 智能体，或任何声明了skillsDir/skillsSharedStore的自定义 Agent）并不通过 ACP 广播技能命令；diskSkillTrigger返回null导致：useAgentSkills传入null，前端不会请求磁盘技能列表，availableSkills永远为空；/时无法弹出磁盘技能候选菜单；/smart-dispatch）无法被识别为技能引用徽标（Token Badge）；修复方案 (Solution)
src/components/chat/composer/invocation-reference.ts中引入isCustomAgentType，将自定义 ACP 智能体纳入diskSkillTrigger的/触发范围。skillsForMenu自身已有对commands的去重过滤机制，若某自定义 Agent 同时存在同名命令也会被安全过滤。acp_list_agent_skills返回supported: false，useAgentSkills自然保持空列表，完全平稳兼容。invocation-reference.test.ts，移除将custom:x视为 null 的断言，补充自定义 ACP 智能体触发/技能的单测用例。message-input.test.tsx补充针对自定义 ACP 智能体斜杠菜单下列出磁盘技能的集成测试。验证情况 (Verification)
src/components/chat/composer/invocation-reference.test.ts(23 passed)src/components/chat/message-input.test.tsx(118 passed)