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 metago-ai/metagolifeform --skill metago-self-eyesgit clone --depth 1 https://github.com/metago-ai/metagolifeformWrote 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/metago-ai/metagolifeform/metago-self-eyes)<a href="https://agentmods.dev/skills/metago-ai/metagolifeform/metago-self-eyes"><img src="https://agentmods.dev/badge/skills/metago-ai/metagolifeform/metago-self-eyes/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/metago-ai/metagolifeform/metago-self-eyes"><img src="https://agentmods.dev/badge/skills/metago-ai/metagolifeform/metago-self-eyes.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 | $0.00149 | $0.04168 |
| Opus 5.5 | $0.00060 | $0.01667 |
| Sonnet 5.5 | $0.00030 | $0.00834 |
| Haiku 4.5 | $0.00015 | $0.00417 |
Grade A, and why
metago-self-eyes 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 — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
自见协议（Self-Eyes Protocol）
一句话铁律：没看过产物，就没有资格说"完成"。 第二条铁律：没跑过自测，就没有资格说"审计器可信"。
0. 为什么需要这份协议
我是作者，天然偏心。没有外部画面证据时，我会用"逻辑上应该没问题"替自己圆过去—— 于是出现：需求听懂了、代码也按需求写了，交付出去仍然排版塌、文字截断、图文不符、手机上一团乱。
测试通过 ≠ 看起来对。tsc 0 错误 ≠ 用户满意。HTTP 200 ≠ 页面没塌。
1. 四步闭环（缺一不可）
① 渲染 把产物变成"用户会看到的像素"（截图/抽帧/页面照片）
② 看 用 read_image 真的把那张图读进来看（不是看 DOM、不是看代码）
③ 挑 对照本技能的分类型清单，逐项判定 PASS/FAIL，指出位置
④ 修+复看 改完必须重新渲染再看一次，直到清单全绿
禁止跳过 ②。 截了图不读图 = 没看。
看不清就放大，不要靠猜。 整图 read_image 会被缩放到 ~2600px 宽，局部细节全糊；先用 ffmpeg crop + scale 把局部放大 2 倍再 read_image（ffmpeg 用绝对路径，WinGet 装的形如 C:\Users\<用户>\AppData\Local\Microsoft\WinGet\Packages\Gyan.FFmpeg_...\bin\ffmpeg.exe）：
& '<ffmpeg 绝对路径>' -y -v error -i D:\审计\full.png -vf "crop=470:1850:0:0,scale=940:-1" D:\审计\crop-sidebar.png
crop=宽:高:x:y 取局部，scale=宽:-1 等比放大。左栏、底部输入区、单张卡片看不清时都这么看。
2. 分类型：怎么看
下文所有命令都用绝对路径，可直接复制到任意工作目录执行。node 装在
C:\Program Files\nodejs\node.exe；没有 pwsh（PowerShell 7）时用 Windows PowerShell 5.1：powershell -NoProfile -ExecutionPolicy Bypass -File <脚本>。
2.1 网站 / 软件界面（最高频）
# 本地页面 / 活界面（只读自审：不点按钮、不动用户窗口）
& 'C:\Program Files\nodejs\node.exe' $DSH_HOME\home\skills\metago-self-eyes\scripts\audit-web.mjs --url http://127.0.0.1:<端口> --out D:\审计\<名> --wait 4000 --widths 1440
# 本地 HTML 文件
& 'C:\Program Files\nodejs\node.exe' $DSH_HOME\home\skills\metago-self-eyes\scripts\audit-web.mjs --html D:\path\page.html --out D:\审计\<名>
产出：shots/*.png（1440 / 768 / 390 三个视口）+ report.json（客观问题清单）。
然后必须 read_image 逐张看，重点看：
| # | 看什么 | 失败长什么样 |
|---|---|---|
| 1 | 首屏有没有内容 | 大面积空白 / 只有一个标题孤零零 |
| 2 | 文字有没有被切 | 行尾断字、按钮文字溢出、卡片文字压边 |
| 3 | 对齐与栅格 | 左右边距不一致、卡片高度参差、基线不齐 |
| 4 | 层级与对比 | 次要文字抢主导、浅灰文字看不清、深色块上黑字 |
| 5 | 留白与密度 | 挤成一坨 / 空得发慌 |
| 6 | 遮挡与溢出 | 弹层被裁、下拉被容器切掉、横向滚动条 |
| 7 | 空状态与长文本 | 没数据时白板、长标题撑破布局 |
| 8 | 真实内容 vs 占位 | 还是 Lorem / 示例数字 / TODO |
| 9 | 与需求逐条对照 | 需求说 A，界面给 B |
| 10 | 破图与控制台报错 | 图片裂开、红色报错 |
交互必须动手看：hover / 点击 / 打开弹窗 / 提交空表单 / 滚动到底 —— 每个动作后再看一次截图。
What ships with it
23 files 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.
- .selftest-out/clean-page/report.json 1.2 KB
- .selftest-out/clean-page/shots/1440-full.png 23 KB
- .selftest-out/clean-page/shots/1440-viewport.png 23 KB
- .selftest-out/clean-page/shots/390-full.png 20 KB
- .selftest-out/clean-page/shots/390-viewport.png 20 KB
- .selftest-out/empty-page/report.json 1.5 KB
- .selftest-out/empty-page/shots/1440-full.png 5.7 KB
- .selftest-out/empty-page/shots/1440-viewport.png 5.7 KB
- .selftest-out/empty-page/shots/390-full.png 2.7 KB
- .selftest-out/empty-page/shots/390-viewport.png 2.7 KB
- .selftest-out/selftest-page/report.json 5.8 KB
- .selftest-out/selftest-page/shots/1440-full.png 38 KB
- .selftest-out/selftest-page/shots/1440-viewport.png 38 KB
- .selftest-out/selftest-page/shots/390-full.png 38 KB
- .selftest-out/selftest-page/shots/390-viewport.png 33 KB
- fixtures/clean-page.html 1.1 KB
- fixtures/empty-page.html 118 B
- scripts/audit-image.mjs 3.9 KB runs code
- scripts/audit-video.mjs 4.8 KB runs code
- scripts/audit-web.mjs 11 KB runs code
- scripts/capture-window.ps1 4.0 KB runs code
- scripts/selftest-page.html 2.9 KB
- scripts/selftest.mjs 2.9 KB runs code
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 · 220 lines · 149 tokens per session scan A 0676e540e397
metago-self-eyes is a skill published in the GitHub repository metago-ai/metagolifeform (6 stars, last pushed yesterday), licensed MIT. It adds 149 tokens to every session and 4,168 once invoked, about $0.0006 per session on Opus 5.5. 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 2026-11-06.
metago-self-eyes compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| metago-self-eyesmetago-ai/metagolifeform | Skill | 6 | yesterday | MIT | A | 149 | Claude Code, Codex |
| todo-for-ai-agent-runtimetodo-for-ai/todo-for-ai | Skill | 1.2k | 1mo ago | Apache-2.0 | A | 39 | Claude Code, Codex |
| monetization-guidegenapohub/monetization-guide | Skill | 3 | 2mo ago | MIT | A | 121 | Claude Code, Codex |
| wechat-article-scraperlililly123789/wechat-article-scraper | Skill | 1 | 3mo ago | MIT | A | 56 | Claude Code, Codex |
| karpathy-guidelinesVincent-A-Yang/karpathy-skills-anycoding | Skill | 3 | 6mo ago | MIT | A | 47 | Claude Code, Codex |
| prototype-webnexu-io/html-anything | Skill | 9.5k | today | Apache-2.0 | A | 24 | Claude Code, Codex |
| loop-engineeringhuytieu/COG-second-brain | Skill | 1.3k | 17d ago | MIT | A | 63 | Claude Code |