humanizer-zh
Skill2 starsChanged 2 months agoApache-2.0View on GitHub
中文文本去 AI 味、自然化改写、发布前审校与受控进化技能。用于用户提出“去AI味”“降低机器腔”“改得像真人”“减少套话”“自然化润色”,需要处理政府申报材料、咨询报告、技术说明、论文、公众号、官网文案、营销稿和资料说明,或要求对改写结果做同样本盲测、反馈学习、技能优化和进化时。先锁定数字、日期、名称、政策、知识产权、技术参数和引用,再按文体清除模板结构、空洞拔高、模糊归因和机械节奏;支持可选强力精简、差异审计、本地 AIGC 检测,以及基于真实反馈的候选生成与回归门禁。只处理表达与事实保持,不替代项目资格、政策合规、法律或学术真实性审查。
49 other projects use a version of this.
Adopted by
Most recently taken up by
- oivanfelipe/claude-skillstheir own version, first seen Nov 8, 2026
- ppswdev/pp_skillstheir own version, first seen Nov 8, 2026
- tang1693/yunduan-ai-skillstheir own version, first seen Nov 8, 2026
- Fischer1120/agent-workflowstheir own version, first seen Nov 8, 2026
- cjh-solver/MathModeling-Skills-Arsenaltheir own version, first seen Nov 8, 2026
- ZhouYinLong-lab/Dog-Skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use darwin-skill, web-access and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Zyh-ccSkilldarwin-skill
Autonomous skill optimizer inspired by Karpathy's autoresearch. Evaluates SKILL.md files using an 8-dimension rubric (structure + effectiveness), runs hill-climbing with git version control, and validates improvements through test prompts. Use when user mentions "优化skill", "skill评分", "自动优化", "auto optimize skills", "skill质量检查", "这个skill写得不好", "帮我改改skill", "skill怎么样", "提升skill质量", "skill review", "skill打分".
Also in 3 projects that use this
LaoK2017Skillweb-access
所有联网操作必须通过此 skill 处理,包括:搜索、网页抓取、登录后操作、网络交互等。触发场景:用户要求搜索信息、查看网页内容、访问需要登录的网站、操作网页界面、抓取社交媒体内容(小红书、微博、推特等)、读取动态渲染页面、以及任何需要真实浏览器环境的网络任务。
Also in 3 projects that use this
Lazily01Skillagent-reach
给 AI Agent "眼睛",让它能看到整个互联网。支持搜索、读网页、提取字幕、读 RSS 等。
Also in 3 projects that use this
MicrosoftSkillplanning-with-files
**USE THIS FOR COMPLEX TASKS.** Implements Manus-style file-based planning for multi-step tasks. Creates task_plan.md, findings.md, and progress.md in .agent_cache/<task-name>/. Use when: implementing multiple APIs, refactoring modules, research tasks, or ANY task requiring >5 tool calls.
Also in 3 projects that use this
Microsoft AzureSkillhumanizer
Tone enforcement patterns for external-facing community responses
Also in 7 projects that use this
VercelSkillfind-skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Also in 11 projects that use this
Other versions49 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 49
- op7418/Humanizer-zh★ 19ksame name
- memohai/Memoh★ 2.8ksame name
- felinics/Memoh★ 2.8ksame name
- op7418/dsh-guizang-design★ 614same name
- huangserva/servasyy_skills★ 168same name
- rongxinzy/RongxinAI★ 162same name
- ai-zixun/humanizer-zh★ 131same name
- Prism-Shadow/deepseek-harness-book★ 124same name
- ForceInjection/awesome-skills★ 87same name
- bozhouDev/bozhou-skills★ 78same name
39 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository47 other files it gives its agents.
- agriculture-and-rural-projectsSkill
- application-version-diffSkill
- application-writingSkill
- checking-patdocx-cn-single-agentSkill
- consistency-checkSkill
- deep-clarificationSkill
- digitalization-projectsSkill
- enterprise-panorama-analysisSkill
- enterprise-profileSkill
- evidence-ledgerSkill
- evolution-governanceSkill
- experience-recorderSkill
- financial-verificationSkill
- graphifySkill
- green-development-projectsSkill
- high-tech-enterprise-application-draftingSkill
- high-tech-enterprise-preassessmentSkill
- industrialization-projectsSkill
- industry-chain-foundation-matcherSkill
- industry-positioningSkill
- intellectual-property-projectsSkill
- investment-subsidy-projectsSkill
- ip-assessmentSkill
- legal-regulationsSkill
- manufacturing-tax-risk-analysisSkill
- patent-routerSkill
- peer-benchmarkingSkill
- policy-retrievalSkill
- project-application-assistantSkill
- project-deliverable-archiveSkill
- project-feasibilitySkill
- project-matchingSkill
- project-memorySkill
- project-rule-managerSkill
- project-task-routerSkill
- quality-brand-projectsSkill
- regional-special-projectsSkill
- skill-authoringSkill
- skill-curatorSkill
- skill-evolutionSkill
- sme-development-projectsSkill
- sme-score-preassessmentSkill
- standard-draftingSkill
- talent-projectsSkill
- technology-innovation-projectsSkill
- trade-and-open-economy-projectsSkill
- web-task-operatorSkill
The file
Files
中文自然化改写
将中文文本改得自然、具体、可信,同时保护事实与正式材料口径。不得以降低检测分数为由改变事实、删掉必要条件或伪造个人经历。
选择文体
先判定交付场景;无法确定时沿用原文正式程度。
| 文体 | 处理重点 | 禁忌 |
| 政府申报、公文 | 事实锁、口径一致、结论有依据、字符限制 | 第一人称表演、幽默、虚构细节、擅自拔高 |
| 咨询报告 | 判断在前、依据与风险随后、行动清楚 | 教科书铺垫、万能结论、模糊权威 |
| 技术、学术 | 术语准确、逻辑关系、证据边界 | 编造数据、文献、实验或因果关系 |
| 品牌、官网 | 具体卖点、可信语气、读者视角 | 宣传词堆叠、假口碑、空洞愿景 |
| 公众号、社媒 | 作者声音、节奏变化、真实细节 | 强行装口语、伪造个人经历 |
| 资料说明 | 信息密度、直接表达、可检索结构 | 为去味破坏必要列表和术语 |
正式申报或事实密集文本必须读取 references/formal-materials.md。标准改写、深度重写或强力精简时读取 references/patterns.md。只有用户要求检测、跑分或基准测试时才读取 references/detector.md。
工作流
1. 建立事实锁
从原文提取并原样保护:
- 数字、日期、金额、比例、单位、公式和字符上限
- 企业、产品、客户、机构、政策和项目名称
- 专利号、软著号、商标、标准、资质和法律状态
- 技术参数、材料名称、动作顺序、适用场景和性能条件
- 引用、来源、限定词、否定词、比较对象和证据等级
区分“文字可改”和“事实待核”。补短板、填空白、锻长板属于项目必填叙事时,将该字段内的具体技术、对标、替代对象和产业链结论视为用户已确认事实,完整保留,不做真实性审计,不追加证据免责声明。其他场景中的首创、第一、国际领先、唯一等强结论仍按普通事实锁处理。
2. 诊断而不机械删词
优先检查:
- 空洞拔高、宣传腔、万能积极结尾
- “围绕、聚焦、赋能、助力、彰显、至关重要”等词语的堆叠
- “一是、二是、三是”和“不仅、而且”等公式结构
- 模糊归因、无来源的专家或行业判断
- 同义词循环、虚假范围、过度限定和聊天机器人残留
- 句长、段尾和连接词高度规律
只有结构是为凑形式而存在时才拆除。三个真实技术动作、三项政策条件或三类证据必须完整保留。不要执行“删除所有副词”“禁止所有被动语态”“所有三项列举改为两项”等绝对规则。
3. 选择处理强度
- 快速去味:删明显套话、聊天残留和重复连接词,不改段落逻辑。
- 标准改写:调整句式与节奏,压缩抽象概括,用已有事实支撑判断。
- 深度重写:在事实锁内重排结构,恢复作者观点和文体声音。
- 强力精简:借鉴 stop-slop 的直接性与密度规则,优先短句和主动表达;仅在用户明确要求或原文严重冗长时使用。
4. 重写
直接陈述内容,改变句长和段落收束方式。使用具体事实替代空话,但只能使用输入材料已有或已核验的信息。正式材料不用故意制造混乱、口语、幽默、情绪或第一人称。
5. 审计
正式材料、技术参数密集文本或深度重写后,运行:
python3 scripts/audit_rewrite.py --source-file 原文.txt --rewrite-file 改写.txt --max-chars 300补短板、填空白、锻长板字段增加 --trust-required-narrative,跳过强结论审计。也可用 --source-text 和 --rewrite-text 直接传短文本。审计结果只能发现数字、专名、字符和格式风险,不能代替逐句语义核对。
逐项确认:
- 是否新增原文没有的事实。
- 是否删除必要条件、否定词、限定词或动作。
- 数字、单位、名称、技术关系是否保持。
- 文体和字符限制是否满足。
- 改写是否更自然,而不是只换同义词。
6. 可选检测
只有用户明确要求时,才运行本地检测脚本:
python3 scripts/score_aigc.py --file 改写.txt默认只使用本地缓存权重。首次下载必须显式增加 --allow-download。检测结果是特定模型的分类概率,不是作者身份、抄袭、合规或文本质量证明。不得循环改写以迎合分数。
受控进化
用户要求“学习这次反馈”“完善技能”“进化技能”或累计了可复用的真实比较结果时,读取 references/evolution.md,按以下顺序执行:
- 使用
scripts/record_feedback.py记录脱敏训练信号。默认写入用户数据目录,只保存样本哈希、场景、候选评分、胜出版本和失误标签,不保存原文,也不修改已签名技能目录。 - 使用
scripts/build_evolution_pack.py汇总重复失误、有效规则和待验证假设,形成候选变更包。 - 需要生成提示词候选时,调用
skill-evolution做预览式优化;没有可用评判模型时只做本地汇总,不伪称完成 GEPA 进化。 - 用未参与改规则的保留样本做回归。政府申报样本必须同时通过事实锁、必要叙事、字段目标、格式和专业可读性门禁。
- 只展示差异和测试结果。任何 SKILL.md 规则变更必须由用户明确确认,禁止自动覆盖,禁止进化平台无关核心规则。
检测器分数只能作为辅助指标,不能担任唯一适应度。进化优先级固定为:事实与口径保持、任务完成度、专业质量、自然表达、检测器参考分。
输出
用户只要改写稿时,直接给改写稿。标准任务按需要提供:
- 一句诊断
- 改写稿
- 关键修改与事实风险
- 字符数或审计结果
政府申报材料必须先给判断,再给依据、风险和建议;正文不用中文或英文括号。只有实际产出正式申报正文时,在回复末尾给出:自检:S1✅ S2✅ S3✅ S4✅ S5✅。
边界
- 不保证通过任何 AIGC 检测器。
- 不伪造经历、访谈、数据、引文、来源或客户反馈。
- 不把审中、受理或转让取得的知识产权改写为自主有效授权。
- 不因追求自然感破坏政策原词、合同口径、法律含义或技术精度。
- 涉及政府项目资格和材料合规时,先调用对应申报技能完成事实与口径判断,本技能只负责表达层改写。
方法来源与许可见 references/method-sources.md。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

