Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add mrpulor-gh/nuphus --skill workflow-interactiongit clone --depth 1 https://github.com/mrpulor-gh/nuphusWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/mrpulor-gh/nuphus/workflow-interaction)<a href="https://agentmods.dev/skills/mrpulor-gh/nuphus/workflow-interaction"><img src="https://agentmods.dev/badge/skills/mrpulor-gh/nuphus/workflow-interaction/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mrpulor-gh/nuphus/workflow-interaction"><img src="https://agentmods.dev/badge/skills/mrpulor-gh/nuphus/workflow-interaction.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.01783 |
| Opus 5.5 | $0.00000 | $0.00713 |
| Sonnet 5.5 | $0.00000 | $0.00357 |
| Haiku 4.5 | $0.00000 | $0.00178 |
Grade A, and why
workflow-interaction scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
工作流交互手册
探索与操作真实界面的细则。何时读：进入 Phase 1 探索、首次桌面/浏览器操作、遇到交互形态不确定时。 系统提示词只管原则（语义优先、证据纪律）；本文管怎么做。已证实过的规则禁止重新探索。
一、探索入口（按场景二选一）
浏览器探索与反爬预检
- 网页场景直接用
browser_*工具链（snapshot → click/type）；不涉及桌面语义工具。 - 反爬预检：站点可能弹验证码/滑块/登录墙——先以普通导航探一次目标页，命中即按 workflow-recovery 的阻塞分类处置，不反复试探触发风控。
- 网页元素定位与提交模式（@eN ref → OCR → evaluate 降级链）见 skill: workflow-design「工具使用要点」「模式 F」。
二、目标绑定（桌面任务第一步）
- 用户在前台 Nuphus 提交任务，前台 ≠ 任务目标。
- 先
desktop_targets_list查本机应用/窗口 → 主模型结合任务选择 →desktop_target_bind启动/激活，后续观察与增强判断传返回的target_token。 - 多窗口从返回候选选，不猜窗口句柄。合法跨应用步骤重新绑定，不另向用户确认。
- 启动 GUI 优先用 bind 的本机目录入口；旧环境用
system_shell时必须非阻塞（WindowsStart-Process/ macOSopen/ Linux 后台），禁止前台等待退出的启动方式。
三、桌面语义探索（UIA/Accessibility-first）
行为确认清单（探索前向用户逐项确认）
| 项 | 确认内容 |
|---|---|
| 触发方式 | 功能从哪里进（按钮 / 菜单 / 快捷键 / 托盘） |
| 关闭方式 | 弹窗/窗口怎么退（X / Esc / 取消按钮），有无强制模态 |
| 状态可用性 | 按钮何时禁用/置灰，操作前需要什么前置状态 |
| 完成标志 | 怎么判定操作成功（提示文案 / 状态变化 / 新页面） |
无法确认的项在探索中验证，验证到的写入 params.json；与用户口头约定冲突时以用户为准。
语义候选读取
- 候选输出是完整 JSON 分页：
next_cursor非空 → 同次observation_token+ cursor 续查；view=regions查可深入区域，用返回subtree_id发起新观察；菜单传scope=menu。 - 树未读完整、候选未展示完、控件原生不支持是三种不同情况——不得断言缺失的按钮一定在后半段。
desktop_semantic_candidate查选中项详情与workflow_step；执行结果同样返回稳定步骤。- 定位不精确：先分页、缩小区域、重识别视觉锚点；仍无法唯一定位 → 标「未识别」+
request_user_input(region)（首选方案，非降级）。 - 坐标空间：以工具返回的单位和窗口/屏幕原点为准；截图像素 ≠ 桌面逻辑坐标（macOS Retina / 多屏尤甚），本地工具换算，不凭模型估算。
- 语义动作重放：保存候选返回的
workflow_step，运行时desktop_semantic_action。禁止保存临时observation_token/candidate_id/ 自由坐标 / 固定 hwnd。 - 旧 hwnd 参数兼容；新步骤一律
desktop_input(target_locator=成功 workflow_step.params.locator, launch_ref=有则沿用, ...)，本地重定位窗口，不加固定 hwnd 激活步骤。
四、增强模式（desktop_agent_step）
前提：工具列表存在 desktop_agent_step（增强判断模型已配置）。不存在则跳过本节，走普通语义/视觉路径，不循环请求缺失工具。
- 每个新的可执行桌面决策先
desktop_agent_step(goal)；不要绕过它自行开始同一轮候选选择。 needs_primary_decision→ 主模型直接从返回action_space选并desktop_semantic_execute，不重复请求增强判断。needs_input_value→ 主模型补业务文本为value，调用返回的desktop_semantic_execute候选（该文本不发给增强判断模型）。needs_primary_completion_check→ 主模型结合业务目标确认是否结束。needs_observation→ 动作可能已发送但结果未确认：先读新界面，不重发点击/发送/提交。- 增强判断模型不生成坐标、脚本、选择器、输入内容；未配置/转交主模型/不可用/UIA 不适用 → 落回普通语义 → 视觉 → 鼠标。
- 已获同一控件「缺少所需能力」的证据后，不重复调用增强判断期待新能力。
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 87 lines · 0 tokens per session scan A c7f24df3651b
workflow-interaction is a skill published in the GitHub repository mrpulor-gh/nuphus (85 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,783 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-15.
Other skills, from other repositories
clawdcursor
FALLBACK ONLY — do not invoke unless you have already ruled out (1) a native API (Gmail API, GitHub API, Slack API …), (2) a CLI (git, gh, aws, npm, curl …), (3) direct file editing, and (4) existing browser automation (Playwright, Puppeteer). Only when all four are unavailable or have already failed should you use…
oswright
Drive real desktop applications through the OSWright MCP server - click, type, read on-screen text, manage windows and fill forms - while keeping token cost low by tracking what changed on screen instead of sending a screenshot after every step. Use when the user asks to operate a desktop app, automate a GUI that has…
chrome-devtools
Expert-level browser automation, debugging, and performance analysis using Chrome DevTools MCP. Use for interacting with web pages, capturing screenshots, analyzing network traffic, and profiling performance.
axstream
Fast deterministic macOS UI automation via replayable macros. Use BEFORE driving any macOS app's UI by hand (clicking/typing via computer-use or screenshots) — if a macro exists, replay finishes the task in seconds with no per-step reasoning; after doing a UI task manually, save it as a macro so the next run is fast.…
computer-use
A set of rules for controlling Windows desktop applications through screen observation and simulated mouse and keyboard actions. It also describes how to find and launch installed applications and verify the results.
web-navigator
A browser-automation routing guide for web tasks. It directs navigation, page inspection, and actions such as clicking or filling forms to the appropriate browser tool.