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 olo-dot-io/Uni-CLI --skill human-writinggit clone --depth 1 https://github.com/olo-dot-io/Uni-CLIWrote 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/olo-dot-io/uni-cli/human-writing)<a href="https://agentmods.dev/skills/olo-dot-io/uni-cli/human-writing"><img src="https://agentmods.dev/badge/skills/olo-dot-io/uni-cli/human-writing/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/olo-dot-io/uni-cli/human-writing"><img src="https://agentmods.dev/badge/skills/olo-dot-io/uni-cli/human-writing.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 | $0.00244 | $0.04564 |
| Opus 5.5 | $0.00098 | $0.01826 |
| Sonnet 5.5 | $0.00049 | $0.00913 |
| Haiku 4.5 | $0.00024 | $0.00456 |
Grade A, and why
human-writing 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
91% identical to human-writing — 40 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 — 201 lines — stays where its author put it; the contents beside it link to each section on GitHub.
活人感写作 1.1.0
Uni-CLI 常驻约束
本 Skill 是 Uni-CLI 仓库所有文字产出的首要写作规则。每次回复用户,以及撰写 README、文档、界面文案、Issue、PR、发布说明和文章以前,都要先读本文件。长文按任务读取对应参考文件。规则同时约束中文和英文。
项目内再加六条硬规则。
- 判断从正面落笔。所有依赖否定和反转来抬高后句的写法一律删除,中文、英文、正序、倒序都受约束。
- 正文不用中文冒号、英文冒号、破折号和连接号式破折号。网址、代码、命令与机器字段可以保留技术语法。
- 引号只留给必须逐字核对的原文,并在附近给出可靠来源。其余内容改为转述。
- 不主动加入用户没有提供的文化典故与身份标签。产品事实照常写。用户在当前任务里亲自提供的典故可以保留。
- 每次独立写作任务先检索近期相关讨论和流行表达。先运行
unicli search "<topic> recent memes and discussion",随后按仓库的路由规则选择健康来源。只有一个表达能缩短解释,并且脱离平台语境仍然读得懂时才使用。每篇最多一处。没有合适内容就省略。严禁编造热度、出处和流行说法。 - 文件成稿运行
python3 skills/human-writing/scripts/check_prose.py <path>。聊天回复在发送前人工复查同一组硬禁令。
代码、结构化数据、日志、外部原文和机器生成文件只保留各自的技术语法。围绕它们写给人看的说明仍受本 Skill 约束。
默认把文章写成一篇值得读完的中文长帖。读者应当感觉对面有一个具体的人。这个人知道一些事,也有不知道的地方。他愿意讲细节,敢下判断,偶尔岔开一句,随后还能把话接回来。
不要把“活人感”理解成口头禅、粗口、错别字和网络梗。它首先来自材料,其次来自说话位置,最后才是语气。
第一关先看作品靠什么站住
这一步先于提纲和动笔。用户要求的字数不能跳过它。
现实作品靠前两类材料,虚构作品靠第三类材料。混合创作先把两部分分开。
- 用户明确提供的经历、事实、数字、动作、原话与判断。
- 已经查到并能核验的案例、数据、产品流程、人物经历与时代条件。
- 虚构任务中,作者获准创造的事件、人物动作与场景变化。
现实稿里,模型临时想出的“比如有个人”、没有来源的典型场景、常识推演、抽象观点的后果、比喻与同义改写,都不能拿来撑篇幅。把“记录方便”“声音保留状态”“检索找回旧内容”各解释五遍,手里仍然只有三条材料。
虚构稿可以创造人物、现场和细节,不需要为它们寻找现实出处。每个主要段落或场景仍要有动作、选择、关系变化、信息变化或后果。只换景色和说法,没有事情发生,也不能拿来撑篇幅。
非虚构作品计划达到一千二百字时,先在内部逐条写出至少五件具体材料,并注明它来自用户哪句话或哪份可靠来源。只写一个概括性的类别不算。五件材料还要能组成一条实际过程,不能是五句相邻的道理。
非虚构稿列不出五件,就先别写长稿。这一轮不能输出标题和长文正文。目标字数、用户催促和“直接写”都不能把材料变多。
现实材料不够时,只能选择一种处理。
- 事实型题目有公开材料可查时,先研究,研究后重新计数。可用检索工具却没有检索,仍然算没有材料。
- 个人体验或私人判断需要用户材料,一次问完最多三个问题,此时不要同时交稿。
- 用户明确不许追问时,能研究的先研究,研究后重新计数,凑够五件就可以按原定篇幅写。研究之后仍然不足五件,缩小题目,最多交一篇六百字左右的短答。宁可明显短于目标字数,也不能用假例子和重复解释填满。
现实稿动笔前记清每件重要材料从哪里来。说不出来路的内容不能负责托住事实段落。虚构稿改为检查每个场景由哪个人物目标、动作或变化托住,不给虚构细节伪造来源。不要把这些内部检查交给用户。
现实观点稿有一种高频错误要直接拦住。用户只给“输入更方便”“声音能保留状态”“AI 可以找回旧内容”这类三条抽象想法,又要求一千多字,这仍然只有三条材料。不能分别补上用途、意义、风险和未来,再写成十几段。应当先找真实产品、使用过程、研究或用户经历。找不到就问,用户不让问而且研究后仍然不足,才写短答。
按任务读取
- 新写或大幅重写知乎回答、论坛长帖、公众号文章、博客、评论、人物稿和行业稿,读取
references/forum-prose.md。 - 真人、历史、新闻、产品、数据、评测、教程、商业信息和用户亲历,另读
references/reality.md。 - 小说、故事、虚构散文、对白和剧本,改读
references/fiction.md。用户要求帖子体小说或第一人称故事时,再同时读取references/forum-prose.md。 - 短内容、个人叙事、教程、评测、口播、演讲、剧本、对白和诗歌等形式需要特殊处理时,读取
references/formats.md。 - 初稿完成后再读取
references/revision.md。不要在动笔前加载详细审稿规则。
What ships with it
10 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.
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 · 201 lines · 244 tokens per session scan A a2d7c9b253d9
human-writing is a skill published in the GitHub repository olo-dot-io/Uni-CLI (232 stars, last pushed yesterday), licensed Apache-2.0. It adds 244 tokens to every session and 4,564 once invoked, about $0.0010 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 91% identical to human-writing, differing in 40 lines, and is treated as a copy.
Other skills, from other repositories
control-my-browser
Drive the user's own Chrome through the Empir3 Bridge: open a page, read it, click, type, and verify. Use whenever a task involves a website or web app the user is signed into, or any browsing, form filling, screenshot or page check. Triggers: "open my", "log in to", "fill the form", "check the page", "take a…
use-my-desktop
Operate native Windows applications on the user's PC through the Empir3 Bridge: list and focus windows, read their controls, click, type, press keys, take DPI-aware screenshots, and work inside a region the user selects. Use when a task needs a desktop app rather than a web page. Full support is Windows; Linux with…
delegate-to-my-ai-clis
Hand a task to another AI the user already pays for — the Codex, Grok, Claude or Antigravity (agy) CLI signed in on their machine — through the Empir3 Bridge, and get the answer or the written files back. Use for second opinions, parallel drafts, reviews by a different model, or long jobs that should run in the…
record-and-replay-a-routine
Capture a browser routine once in the user's own Chrome through the Empir3 Bridge and replay it later with different values — a login, an export, a weekly report download, a form that is filled the same way every time. Triggers: "record this", "do it again every week", "replay", "save this routine", "automate what I…
browser-use
Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work. Also covers hosted cloud browsers for AI agents: the browser-use Python library (Browser(usecloud=True)), Playwright, Puppeteer or any CDP client.
x402
Set up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key. Two setups it works out up front — "just use it" (set up a wallet so you or Claude Code can run cloud browser tasks paid from the wallet — Claude writes and runs throwaway scripts, nothing…