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 baimaomaomao556/eagleeye-mcp --skill eagleeyegit clone --depth 1 https://github.com/baimaomaomao556/eagleeye-mcpWrote 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/baimaomaomao556/eagleeye-mcp/eagleeye)<a href="https://agentmods.dev/skills/baimaomaomao556/eagleeye-mcp/eagleeye"><img src="https://agentmods.dev/badge/skills/baimaomaomao556/eagleeye-mcp/eagleeye/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/baimaomaomao556/eagleeye-mcp/eagleeye"><img src="https://agentmods.dev/badge/skills/baimaomaomao556/eagleeye-mcp/eagleeye.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 | $0.00117 | $0.01218 |
| Opus 5.5 | $0.00047 | $0.00487 |
| Sonnet 5.5 | $0.00023 | $0.00244 |
| Haiku 4.5 | $0.00012 | $0.00122 |
Grade A, and why
eagleeye 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 1mo 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.
This is a copy
86% identical to eagleeye — 7 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
EagleEye — DSH / Agent Skills 薄包装
这不是第二个 glance/ground CLI。能力在 MCP 工具 eagleeye_* 上。
本文件只规定:何时用、何时不用、先调哪一层。
仓库计划账号:baimaomaomao556。发布后安装 skill:
npx skills add baimaomaomao556/eagleeye-mcp --skill eagleeye
仍须在 Host 里接入 MCP Server(见仓库 docs/HOSTS.md);本 skill 不替代 stdio 进程。
先看工具列表
- 对话里已经出现
eagleeye_capture_screen等工具 → 直接调用,不要自己写 Pillow/Tesseract,也不要改去跑 modlens/vision-tools 的脚本。 - 没有这些工具 → 停下来告诉用户:需要按仓库
docs/HOSTS.md把 EagleEye 以 stdio MCP 接入当前 Host。不要假装已经截到了屏。
不要用 EagleEye 的时候
- 你已经能看见用户贴的图(原生多模态)→ 自己看,不要再截屏。
- 任务只是「读这张已有图片里的字/布局,输出一份 JSON」→ modlens。
- 任务是长截图 OCR、把界面复原成 HTML、把图标描成 SVG → vision-tools。
- 当前模型被用户规则禁止使用外挂视觉 → 遵守用户规则。
要用 EagleEye 的时候
- 需要当前桌面/指定窗口的真实像素(游戏 UI、前端页面、桌面应用)。
- 需要像素级事实:颜色、偏移、模板是否一致、回归有没有坏。
- 需要把一次捕获的图裁开、放大、再量,并且坐标还能追回屏幕。
- 需要量化「画质/审美」分数、做前后对比(改 UI/美术后再截一张评分,看有没有变好)→
score_image(task=aesthetic|quality,模型qrealign-mini默认 /musiq-ava/nima;source_type=model_inference, 非校准概率,用于粗排与对比)。
Windows 上可以用 list_windows / capture_window / capture_window_graphics。
其它平台这些工具会返回 PLATFORM_UNSUPPORTED:改用 capture_screen 或 load_image。
调用顺序
- Observe:需要等界面出现或加载完成时先
wait_for_visual,再list_windows/list_monitors(需要时)→capture_screen/capture_window/load_image。命中后的image_ref可直接给后续工具,不要再截一张。 - Locate:已知坐标 →
crop_image;已知颜色 →find_color;已知模板 →template_match;已知文字 →locate_text/ocr;只知道「是什么」→locate_visual_element(推断,必须复核)。 - Verify:颜色/尺寸用
get_pixel/sample_region/measure_region;差异用compare_images;pass/fail 用compare_with_reference或visual_regression(动态区域配ignore_regions);画质/审美评分用score_image。
优先级:算法视觉 > OCR > 视觉模型 > 组合工具。能算不算,能测不猜。
region 参数可传对象 {x,y,width,height},也兼容 [x,y,w,h] 与 {w,h} 别名。
小尺寸裁剪/放大图(最长边 <512px)送 analyze_image 等前,server 会自动放大预处理
(EAGLEEYE_VISION_MIN_DIM 可配);若仍 VISION_CALL_FAILED,改用更大范围裁剪或直接对整图提问。
capture_window = 用户屏幕上实际看到的窗口(可能被遮挡)。
capture_window_graphics = 应用渲染表面(可后台),不代表用户当前看见的画面。
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.
- 1mo ago First seen · 71 lines · 117 tokens per session scan A 4af7b4c4f88b
eagleeye is a skill published in the GitHub repository baimaomaomao556/eagleeye-mcp (0 stars, last pushed 2mo ago), licensed MIT. It adds 117 tokens to every session and 1,218 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 86% identical to eagleeye, differing in 7 lines, and is treated as a copy.
Other skills, from other repositories
browser-edge-cases
SOP for debugging browser automation failures on complex websites. Use when browser tools fail on specific sites like LinkedIn, Twitter/X, SPAs, or sites with Shadow DOM.
browserwing-admin
Manage and operate BrowserWing — an intelligent browser automation platform. Install dependencies, configure LLM, create/manage/execute automation scripts, use AI-driven exploration to generate scripts, browse the script marketplace, and troubleshoot issues.
browser-skill
Automate the user's logged-in Chromium browser: read pages, fill forms, scrape data, operate tabs, test a UI, or debug a website. Requires the bsk CLI and browser extension.
playwright-interactive
Persistent browser and Electron interaction through jsrepl for fast iterative UI debugging.
chrome-devtools
Browser debugging, performance profiling, and automation via Chrome DevTools MCP. Use when user says "debug this page", "take a screenshot", "check network requests", "profile performance", "inspect console errors", or "analyze page load". Do NOT use for full E2E test suites (use playwright-skill) or non-browser…
phone-harness
Use when an agent needs to control a real phone — iPhone via Mac iPhone Mirroring or Android via adb (open apps, tap, type, swipe, OCR the screen). Android path runs on this Linux workstation (adb), targets phonefarm devices. iOS path needs macOS + iPhone Mirroring. Use for phonefarm device debugging, OTP/2FA flows…