菜单
Runs DeepSeek Harness headlessly for bounded repository tasks. Use when the user asks to delegate analysis or implementation to DeepSeek Harness.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/CherryHQ/cherry-studio --skill code-mate-deepseek-harness命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when Cherry Studio 用户希望报告、提交或整理 BUG、UI/UX 问题或功能建议，但未明确要求创建 GitHub Issue。
Runs Antigravity CLI headlessly for repository analysis and coding tasks. Use when the user asks to delegate work to Antigravity CLI or compare its result with another coding agent.
Prepare a new release by collecting commits, generating bilingual release notes, updating version files, and creating a release branch. Use when asked to prepare/create a release, bump version, or run `/prepare-release`.
基于 SOC 职业分类
正在显示 SKILL.md
| name | code-mate-deepseek-harness |
| description | Runs DeepSeek Harness headlessly for bounded repository tasks. Use when the user asks to delegate analysis or implementation to DeepSeek Harness. |
command -v dsh. If it is missing, stop and ask the user to install DeepSeek Harness in Code Mate.dsh --profile headless "<prompt>"
Pass the prompt as one quoted argument. Treat stdout as final text, not JSON. Exit zero means success; exit one means the task failed or did not complete. Never start the interactive UI or a login flow.
If DSH reports a missing provider, model, or API configuration, stop and ask the user to configure DeepSeek Harness in Code Mate. Never request, read, print, or copy credentials.
Headless DSH can write to the workspace and creates a persistent session. For read-only work, explicitly tell it not to modify files and inspect the repository diff afterward. Allow modifications only when the user explicitly requested workspace changes, and restrict the working directory to the intended project.
Example: ask DSH to diagnose a test failure without editing, run it in the repository, confirm exit zero and a clean diff, then summarize its final text.