Back to Details

yo-slop

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Yọ Slop

Yọ Slop

Make supplied prose clearer and less machine-shaped while preserving its contract. Keep exact identifiers, quotations, facts, citations, schemas, required formats, accepted terminology, and the owning artifact's decisions.
This is a direct or conditional cleanup outcome, not a mandatory tail stage after 
technical-writing
or another writing owner. Use it when cleanup/pruning is explicitly requested or a material filler, AI-tell, repetition, vague-abstraction, or instruction-noise problem remains in otherwise settled prose.
在保留文本核心信息的前提下，让提供的文本更清晰、更具自然感（而非机器生成的生硬风格）。保留精确的标识符、引用内容、事实、引用来源、架构、要求格式、通用术语以及所属文档的既定决策。
这是一个直接或有条件的清理操作，并非
technical-writing
或其他写作流程后的强制收尾步骤。仅当明确要求进行清理/精简，或已敲定的文本中仍存在大量冗余内容、AI痕迹、重复表述、模糊抽象或指令噪声问题时，才使用此工具。

Choose the pass

选择处理模式

Use the normal cleanup pass for local clarity and voice problems. When the request explicitly asks to shorten prose, reduce verbosity, remove repetition, or lower instruction load, use 
prune
and read prune prose. Do not infer 
prune
from an ordinary editing or final-pass request.
针对局部清晰度或语气问题，使用常规清理模式。当请求明确要求缩短文本、减少冗余、消除重复或降低指令负载时，请使用
prune
模式，并阅读prune prose文档。不要从普通编辑或最终审核请求中推断出需要使用
prune
模式。

Establish the pass

确定处理前提

Identify the reader, artifact owner, purpose, language and locale, intended voice, and text that must remain exact. If these are unclear, fix only unambiguous filler and ask before a change that could alter meaning or tone.
Use the owning skill for content structure, evidence, technical truth, accessibility, localization, or acceptance. 
technical-writing
owns human-facing developer documentation, technical communication, and applicable artifact copy. 
ko-skill
owns skill behavior and proof. Yọ Slop owns cleanup and explicit pruning of settled prose, not structural or factual changes.
明确读者对象、文档所有者、写作目的、语言及区域、预期语气，以及必须完全保留的文本内容。若这些信息不明确，仅修正明确的冗余内容，在进行可能改变语义或语气的修改前需先询问相关人员。
内容结构、证据、技术真实性、可访问性、本地化或验收标准由对应的Skill负责：
technical-writing
负责面向人类的开发者文档、技术沟通内容及相关文档副本；
ko-skill
负责Skill的行为逻辑及验证；Yọ Slop负责已敲定文本的清理和明确要求的精简工作，不负责结构或事实层面的修改。

Match the reader

匹配读者需求

For human-facing prose, sound direct and specific. Vary sentence length naturally. Use a point of view only when the artifact permits one. Keep reference material neutral. Do not add deliberate mess, personality, or opinion to simulate humanity.
For agent-facing prose, optimize execution rather than personality:
  • Preserve the existing control flow, triggers, branches, completion criteria, commands, identifiers, statuses, output fields, and authority boundaries.
  • Make an existing action, condition, or decision easier to parse only when the wording change is behaviorally equivalent.
  • Retain context pointers and the cases they name. Flag a weak or missing branch for the artifact owner instead of changing disclosure behavior.
  • Flag apparent generic defaults, stale environment caches, or rationale that may not change judgment instead of removing them.
  • Preserve prohibitions and deliberate repetition unless the artifact owner confirms that they are redundant.
Report an agent-facing ambiguity when fixing it would add, remove, reorder, or reinterpret behavior. The artifact owner decides the correction.
对于面向人类的文本，表述要直接具体。自然变换句子长度。仅当文档允许时才使用特定视角。参考资料保持中立。不要刻意添加杂乱内容、个性元素或观点来模拟人类风格。
对于面向Agent的文本，优先优化执行效率而非个性：
  • 保留现有的控制流程、触发条件、分支逻辑、完成标准、命令、标识符、状态、输出字段及权限边界。
  • 仅当措辞修改不会改变行为逻辑时，才简化现有操作、条件或决策的表述，使其更易解析。
  • 保留上下文指针及其对应的场景。若发现分支逻辑薄弱或缺失，需向文档所有者标记问题，而非修改信息披露行为。
  • 若发现明显的通用默认值、过时的环境缓存或可能影响判断的理由，需标记问题，而非直接删除。
  • 保留禁止性条款和刻意的重复内容，除非文档所有者确认这些内容是冗余的。
当修复面向Agent文本中的歧义会导致添加、删除、重新排序或重新解释行为逻辑时，需报告该歧义问题，由文档所有者决定修正方案。

Remove patterns by effect

按问题类型移除不良表述模式

Do not replace words mechanically. Rewrite only when a pattern causes puffery, ambiguity, unsupported certainty, needless load, or a voice mismatch.
  • Inflation and promotion: Remove grand significance claims, notability lists, promotional adjectives, generic challenge-and-triumph framing, and conclusions that add no fact or action.
  • Unsupported attribution: Name the source behind “experts say” or “reports suggest,” qualify the claim from evidence, or remove it.
  • Abstract padding: Replace vague metaphors, nominalizations, copula avoidance, ornamental 
    -ing
    clauses, weak verbs, and unmeasured adverbs with the concrete actor, mechanism, action, or result. Common signals include non-literal “landscape,” “tapestry,” “substrate,” or “vector,” and padded verbs such as “serves as,” “showcases,” “leverages,” or “facilitates.” Keep a word when it is an exact or precise domain term.
  • Formulaic construction: Remove forced threes, false ranges, “not just X but Y” framing, synonym cycling, and comparisons that do not change the point.
  • Filler and hedging: Cut throat-clearing, repeated caveats, stacked modals, and phrases such as “in order to” or “it is important to note.” Keep uncertainty that the evidence requires.
  • Assistant performance: Remove chatbot greetings, sycophantic praise, fake excitement, cutoff disclaimers, and closing offers that do not advance the work.
  • Formatting tells: Reduce decorative bold, inline-header repetition, title case, decorative emoji, and punctuation used as a substitute for sentence structure. Preserve the language, project, and artifact conventions. Punctuation is not an error by itself.
  • Ambiguity: Split sentences that carry multiple instructions or unrelated thoughts. Put conditions before guarded actions, keep modifiers beside what they modify, repeat a noun when a pronoun has multiple possible referents, and name the actor when responsibility matters.
  • Fancy synonyms: Prefer the shortest familiar word that preserves precision. Keep established domain terms and exact API, UI, file, flag, command, and code names.
不要机械替换词汇。仅当某种表述模式导致浮夸、歧义、无依据的确定性、不必要的信息负载或语气不符时，才进行改写。
  • 夸大与宣传： 删除无关的重要性宣称、知名列表、宣传性形容词、通用的“挑战-成功”框架，以及未添加任何事实或行动信息的结论。
  • 无依据的归因： 为“专家表示”或“报告显示”明确来源，根据证据限定表述范围，或直接删除此类内容。
  • 抽象冗余： 将模糊隐喻、名词化结构、避免系动词的表述、装饰性的-ing从句、弱动词及无量化的副词替换为具体的参与者、机制、动作或结果。常见信号包括非字面意义的“landscape”“tapestry”“substrate”或“vector”，以及冗余动词如“serves as”“showcases”“leverages”或“facilitates”。若词汇是精确的领域术语，则保留。
  • 公式化结构： 删除刻意的排比、虚假范围、“不仅X而且Y”的框架、同义词堆砌，以及不改变核心观点的对比表述。
  • 冗余与含糊措辞： 删除开场白、重复的警告、多层模态动词，以及“in order to”或“it is important to note”等短语。保留证据要求的不确定性表述。
  • 助手式表述： 删除聊天机器人问候语、谄媚的赞美、虚假的兴奋感、截断声明，以及对推进工作无帮助的收尾提议。
  • 格式痕迹： 减少装饰性粗体、内联标题重复、标题大小写、装饰性表情符号，以及用于替代句子结构的标点符号。保留语言、项目及文档的约定俗成格式。标点本身并非错误。
  • 歧义表述： 拆分包含多个指令或无关想法的句子。将条件置于受约束的动作之前，让修饰语靠近其修饰的对象，当代词有多个可能指代对象时重复名词，当责任归属重要时明确参与者。
  • 花哨同义词： 优先选择能保留精确性的最短常用词汇。保留已确立的领域术语及精确的API、UI、文件、标志、命令和代码名称。

Verify

验证

Read the revision against the original and ask:
  1. Did any fact, authority, condition, status, citation, identifier, schema, or required field change?
  2. Does every remaining sentence tell this reader something needed to decide, act, verify, or understand?
  3. Does the prose fit the artifact instead of performing a generic “human” or “technical” voice?
  4. For agent-facing text, can the agent identify each trigger, branch, completion condition, and exact output without guessing?
  5. What still makes the text sound generated, vague, or over-structured?
Restore any lost capability or nuance. Return the revised prose first, then report only material meaning, authority, or evidence questions that remain.
对照原文阅读修订后的文本，并思考以下问题：
  1. 事实、权威性、条件、状态、引用来源、标识符、架构或必填字段是否有任何改变？
  2. 每一句保留的内容是否都能为读者提供决策、行动、验证或理解所需的信息？
  3. 文本是否符合所属文档的风格，而非采用通用的“人类”或“技术”语气？
  4. 对于面向Agent的文本，Agent是否无需猜测就能识别每个触发条件、分支逻辑、完成标准及精确输出？
  5. 文本中是否仍存在听起来像生成内容、模糊或过度结构化的部分？
恢复任何丢失的功能或细节。先返回修订后的文本，再仅报告仍存在的重大语义、权威性或证据相关问题。