MateClaw is a pluggable agent runtime that coordinates AI agents, tools, conversations, memory, and workflows. It is designed for personal and multi-user AI assistants, including agents used through chat platforms. Catalogue add-ons provide skills for extending its agent workflows.
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 mateaix/mateclaw --skill deai_humanizegit clone --depth 1 https://github.com/mateaix/mateclawWrote 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/mateaix/mateclaw/deai_humanize)<a href="https://agentmods.dev/skills/mateaix/mateclaw/deai_humanize"><img src="https://agentmods.dev/badge/skills/mateaix/mateclaw/deai_humanize/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/mateaix/mateclaw/deai_humanize"><img src="https://agentmods.dev/badge/skills/mateaix/mateclaw/deai_humanize.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00082 | $0.01617 |
| Opus 5.5 | $0.00033 | $0.00647 |
| Sonnet 5.5 | $0.00016 | $0.00323 |
| Haiku 4.5 | $0.00008 | $0.00162 |
Grade A, and why
deai_humanize 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.
How it starts
The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.
去 AI 化 · 人味改写
把一段"一眼 AI"的中文改写成像真人写的。做法是先量化打分找出 AI 痕迹，再针对性改写，然后复检，循环到达标为止。
重要说明：本技能是一个可解释的写作质量启发式，用来指导改写、提升"人味"。它不保证能骗过任何第三方 AI 检测器，也不以"过检测"为目标——目标是文字读起来自然、具体、有个人声音。
何时触发
用户说"去 AI 味 / 去 AI 化 / 这段太 AI 了 / 帮我改得像人写的 / humanize / 让它不像机器写的"，或在公众号、小红书成文后需要润色时。另外两个创作技能（gzh_article、xhs_note）会用 load_skill deai_humanize 主动调用本技能。
开工前：读取共享人设记忆
先用 recall_structured 取回以下键（snake_case），改写时全程遵守：
content_persona— 用户的人设 / 说话口吻writing_style_gzh— 公众号文风偏好writing_style_xhs— 小红书文风偏好banned_words— 禁用词 / 敏感词signature_blocks— 固定的开场 / 结尾 / 签名段
取不到就用中性默认，不要编造人设。
核心方法：打分 → 改写 → 复检 循环
第 1 步 · 打分
用 run_skill_script 执行本技能的 scripts/ai_trace_score.py，参数是一个 JSON 字符串：
run_skill_script(skill="deai_humanize", script="scripts/ai_trace_score.py",
args='{"text": "<待检测文本>", "platform": "gzh"}')
platform取gzh（公众号，容忍连接词略多、句子偏长）或xhs（小红书，鼓励碎句，句长方差权重更低）。- 脚本返回 JSON：
score：0–100，越高越像 AI。signals：每个维度的{name, value, weight, note}（连接词密度、套话命中、具体度缺失、句长齐整度、清单/破折号滥用、段落齐整度）。spans：命中的具体"扣分片段"（套话、连接词等），改写时优先干掉这些。verdict：human-like|some-ai|strong-ai。
第 2 步 · 改写
若 score 高于阈值（建议 55），针对返回的 signals 里高 value 的维度、以及 spans 列出的片段动手改。改写原则见下。
第 3 步 · 复检并循环
改完再跑一次第 1 步。**循环直到 score ≤ 55 或已改满 3 轮**。每轮都把这一版比上一版降了多少分、还剩哪些 spans` 说清楚。3 轮仍不达标就停手，交出当前最佳版本并说明还剩哪些结构性问题（例如通篇是清单体、缺少真实经历）。
改写原则（AI 痕迹 → 人味）
- 删升华套话：干掉"赋能 / 让我们 / 综上所述 / 在……的今天 / 随着……的发展 / 保驾护航 / 数字化转型 / 打造……新生态"这类词。它们是打分里
cliche_phrases的直接来源。 - 拆连接词骨架：不要"首先……其次……最后……"一条龙。真人靠语义和语气过渡，不靠标签。
- 加具体：塞进真实的数字、时间、地点、人名、场景、亲身经历。抽象的"很重要"换成"上周我因为这事多花了两个小时"。这是
concreteness_deficit维度。 - 句长参差：长短句交替，敢用三五个字的短句，也敢用一个长句。别让每句都一样长（
sentence_burstiness）。 - 第一人称 + 口语化：多用"我 / 我们 / 咱"，用日常说法，允许"其实、说白了、讲真"这类口头语。
- 以具体代抽象：能举例就别下定义，能讲故事就别讲道理。
- 保留不完美感：适度的口语停顿、自我修正、小情绪，比一板一眼的"完美"更像人。
- 少用清单和破折号：不是所有内容都要列点。能用段落叙述的就别拆成 bullet（
list_dash_overuse）。
按平台调口吻
| 维度 | gzh（公众号） | xhs（小红书） |
|---|---|---|
| 句子 | 沉稳、可稍长，逻辑连贯 | 短、碎、跳，一句一断 |
| 语气 | 克制、有分寸、像老友聊深度话题 | 活泼、直给、带情绪和 emoji |
| 结构 | 段落叙述为主 | 痛点共鸣 + 干货碎句 |
| 人称 | "我 / 我们"，偶尔"你" | 大量"我""你""姐妹们" |
What ships with it
2 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 · 97 lines · 82 tokens per session scan A 630b2fad4fd0
deai_humanize is a skill published in the GitHub repository mateaix/mateclaw (1,172 stars, last pushed today), licensed Apache-2.0. It adds 82 tokens to every session and 1,617 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
app
A skill-management module for storing, organising, sharing, and managing reusable skills, including their descriptions, content, tags, and files.
openclaw-security-checker
A security checker for OpenClaw installations, including their configuration, permissions, containers, network access, and logs.
seclens-enterprise-web
Professional web application and API security testing workflows using OWASP Top 10 methodologies.
pentest-recon-attack-surface
White-box attack surface mapping — correlate external scans, browser exploration, and source code into structured endpoint inventory, role architecture, and authorization vulnerability candidates.
pentest-whitebox-code-review
Source code security audit using backward taint analysis, slot type classification, render context verification, and 3-phase parallel review producing an exploitation queue.
pentest-api-deep
Deep OWASP API Security Top 10 testing for REST, GraphQL, gRPC, and WebSocket APIs — BFLA, mass assignment, rate limiting, and unsafe consumption.