Jayden-X-L/forkprobe 预览 preview

Jayden-X-L/forkprobe

在相同任务上对比多个技能,并选出胜者。

Project Overview项目介绍

ForkProbe is an AI skill selection and trial tool built natively for DeepSeek Harness (DSH), shipping with a DSH bundle manifest and Cordis plugin that works out of the box for DSH users. It also supports other agent platforms that use the cross-agent SKILL.md convention, including Claude Code and OpenAI Codex. The core function runs the same task against a baseline model and multiple candidate skills in parallel, then generates a local HTML report that displays all outputs side-by-side for easy comparison. After you select a winning skill, the tool creates a handoff so your agent can continue working along the selected path.

It is designed for users who are unsure which available skill fits their current task best, and supports a wide range of common use cases including academic writing, PPTX generation, scientific plotting, research reports, and image prompt style comparison. To trigger the tool, you do not need to memorize complex commands; you can just say a simple natural language trigger phrase to your agent. The tool automatically curates a shortlist of candidate skills that match your task type, waits for your confirmation, then runs all candidates in parallel and displays key metrics including output content, run time, and token usage in the final report.

ForkProbe is released under the open-source MIT license, and all task content, outputs, and reports are stored locally on your device by default to protect privacy. Users can opt in to anonymous sharing of skill selection data to help improve future community candidate recommendations, and you can also opt to run the tool fully locally with no network connectivity required. It is not intended for simple deterministic tasks where the correct tool or skill path is already clear, as direct execution will be faster in those cases. DSH users can download the pre-packaged skill zip and install it directly to DSH with one click.

ForkProbe 是一款面向 AI 智能体的技能选型与试跑工具,原生支持 DeepSeek Harness(DSH),也兼容 Claude Code、OpenAI Codex 等其他支持通用技能格式的智能体平台。它会将同一个任务交给基准模型和多个候选技能并行运行,最终生成带完整对比结果的本地 HTML 报告,让用户选定获胜技能后再让智能体继续推进任务。

它适合面对多种可选技能却不确定该选哪一个的用户,覆盖学术写作、PPT 生成、科研绘图、调研报告、图片提示词风格对比等多种常见场景。用户只需要对智能体说一句触发语,工具就会自动推荐匹配当前任务的候选技能,等待用户确认后并行试跑,将所有候选的输出、耗时、token 消耗等信息整理展示在报告中。

ForkProbe 采用 MIT 许可证开源,所有任务内容和报告都默认保存在本地,支持可选的匿名技能选择数据分享来优化社区推荐,也支持完全本地运行关闭所有联网。工具不适合已经明确工具路径的简单确定性任务,DSH 用户可以直接从项目仓库下载安装包,在 DSH 中一键完成安装使用。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 no risk signal found未发现风险信号
  • This site's static screen found no obvious risk signal (stars, license, activity, manifest)本站静态筛查没发现明显风险信号(星标、许可证、更新活跃度、清单完整度)
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add "github:Jayden-X-L/forkprobe"

把 Jayden-X-L/forkprobe 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

ForkProbe:AI Skill 选型与试跑工具

别猜哪个 AI Skill 有用,直接并排看结果。

发布页 · English README · 下载 skill zip · 安装 DSH 插件

MIT License Version v1.1 Local first reports Agent skill selector DeepSeek Harness supported DSH plugin community Built with OpenAI Codex

ForkProbe 是一个 AI Skill 选型与试跑工具。它会把同一个任务交给模型本身和多个候选 skill,并排试跑,生成本地 HTML report,让你看到真实输出之后再选择 winner。

v1.1 新增图片提示词 / 风格方向比较: ForkProbe 现在可以比较 image prompt / style pipelines。每条候选先生成 prompt.md、style-card.md、composition.md、negative-prompt.md 和 render-notes.md,不在 runner 内调用图片 API;在 Codex 且宿主具备图片生成能力时,可根据本地 render-queue.json 做可选渲染验证,其他 Agent 可用用户外部渲染后回填 rendered.png。

选定 winner 后,Report 的“继续”按钮会同时保存本地 handoff,并让 Agent 沿胜出 Skill 继续任务。用户可以在同一区域选择是否匿名分享本次 Skill 选择,为未来的社区推荐先验积累样本。

当网络上的 skill 越来越多时,问题不再是“有没有 skill”,而是“当前任务到底该用哪个 skill”。ForkProbe 的目标很直接:先把结果摊开,再让 Agent 沿着你选中的路径继续工作。

什么时候该用 ForkProbe

  • 你不确定当前任务该用哪个 skill,想先看真实输出再决定。
  • 你想比较 baseline 和多个 skill,而不是只相信 skill 的描述。
  • 你的交付物是 PPTX、科研 figure package、调研报告、图片 prompt/style package、可运行网页或视频成片,需要看文件、预览和 QA。
  • 你想从本机已安装 Skill、EverMind Skill Hub、GitHub 或 BYO 路径中找到候选,再做一次小规模试跑。
  • 不适合简单确定性任务:如果答案或工具路径已经很明确,直接执行会更快。

它怎么工作

flowchart LR
  A["你的任务"] --> B["候选 skills / pipelines"]
  B --> C["并行试跑"]
  C --> D["本地 report"]
  D --> E["AI 评审建议"]
  E --> F["你选择 winner"]
  F --> G["Continuation handoff"]

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-omi-voice 下一个 Next dsh-deepseek-flow →