Back to Details

slopbeth

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Slopbeth

Slopbeth

Remove machine-writing tells without sanding away the author's meaning or voice. The target is not "detector-proof" prose; it is dense, specific writing where every sentence carries load and detector results stay dated and tool-specific.
在不削弱作者原意和语气的前提下去除机器写作痕迹。我们的目标并非「规避所有检测工具」的文本，而是打造内容密集、表述具体的写作，每句话都承载有效信息，且检测结果仅针对特定工具且时效性有限。

Workflow

工作流程

  1. Classify the task: rewrite; critique; benchmark; detector-facing validation; or skill maintenance.
  2. Separate the brief from the artifact. Long inputs often mix the material with instructions about it: "the note should keep that texture"; "do not turn this into a lesson"; "the rewrite must not promise that the problem cannot recur". Those sentences address you, not the reader. Do what they ask and leave them out of the output. Reprinting them is the same class of error as inventing content, and preservation and density checks will not catch it, because instruction text is specific, sourced, and dense.
  3. Preserve facts first. Lock named entities; numbers; dates; URLs; citations; quotations; technical claims; explicit uncertainty; and the user's requested stance.
  4. Set the evidence boundary. When the user supplies only vague copy, switch to evidence-bound mode: do not invent or assert product features; dates; people; metrics; workflows; examples; customer facts; or outcome claims. Unsupported claims such as "faster decisions," "better alignment," "reduced friction," "confidence," or "momentum" must become proof gaps, questions, or explicitly attributed claims.
  5. Diagnose clusters, not isolated words. Look for filler; vague significance language; formulaic contrast; promotional inflation; padded lists; generic uplift; actorless claims; summary endings; and ornamental formatting.
  6. Rewrite in this order: preserve claims and constraints; cut scaffolding and inflated abstract nouns; apply Orwell's six rules as generation defaults (short word over long, cut deletable words, active over passive, no printed-cliche metaphor or jargon, but break any rule sooner than write something unclear or graceless); make each sentence carry a claim, example, constraint, image, number, consequence, or argumentative move; match the user's register; remove concrete details that are not sourced or clearly labeled; check for meaning loss, bland-clean prose, formula replacement, and over-editing.
  7. Validate when files or before/after text are available. Use the scripts in 
    scripts/
    for repeatable checks, then apply judgment for meaning, voice, and sentence-load failures.
  8. Output the revised text first for normal rewrite requests. Add a compact note only when it helps explain material changes, preservation risks, or remaining issues.
  1. 任务分类：重写、点评、对标、面向检测工具的验证或技能维护。
  2. 区分需求说明与目标文本：长输入往往会将待处理内容与操作说明混合，例如「这段笔记需保留原有质感」「不要把这个改成教程」「重写时不能承诺问题不会再次发生」。这些句子是给你的指令，而非面向读者的内容。请按要求执行，但不要将它们包含在输出结果中。重复输出这些指令属于与编造内容同类的错误，且内容保留和密度检查无法发现这类错误，因为指令文本本身是具体、有来源且内容密集的。
  3. 优先保留事实：锁定命名实体、数字、日期、URL、引用内容、引文、技术声明、明确的不确定性表述以及用户要求的立场。
  4. 设置证据边界：当用户仅提供模糊文本时，切换到证据绑定模式：不得编造或断言产品功能、日期、人物、指标、工作流程、示例、客户事实或成果声明。诸如「更快的决策」「更好的对齐」「减少摩擦」「提升信心」或「增强势头」这类无依据的声明必须改为证据缺口、问题或明确标注来源的声明。
  5. 诊断集群问题，而非孤立词汇：查找冗余内容、模糊的重要性表述、公式化对比、夸大的宣传语、冗余列表、泛化的提升表述、无主语声明、总结式结尾以及装饰性格式。
  6. 按以下顺序重写：保留声明与约束条件；删除框架性内容和夸大的抽象名词；默认遵循Orwell的六条规则（用短词替代长词、删除冗余词汇、主动语态替代被动语态、避免陈词滥调的隐喻或行话，但为了避免表述不清或生硬，可打破任何规则）；确保每句话都承载一个声明、示例、约束条件、意象、数字、结果或论证步骤；匹配用户的语体风格；移除无来源或未明确标注的具体细节；检查是否存在意义丢失、平淡无奇的文本、公式化替换以及过度编辑的问题。
  7. 当有文件或前后对比文本时进行验证：使用
    scripts/
    目录下的脚本进行可重复检查，然后针对意义、语气和句子信息承载度的问题进行主观判断。
  8. 对于常规重写请求，先输出修订后的文本。仅当需要解释重大修改、保留风险或剩余问题时，添加简洁说明。

Reference routing

参考资料调用

Load only the references needed for the task:
  • references/slop-taxonomy.md
    : thorough diagnosis; red-team review; marker inventory.
  • references/voice-and-preservation.md
    : author samples; technical prose; legal, medical, or financial claims; tone preservation.
  • references/density-and-unsummarizability.md
    : dense prose; stronger argumentation; the user's "unsummarizable" standard.
  • references/writing-system.md
    : generating prose from a positive system; Orwell's six rules; passive-voice reduction; a portable CLAUDE.md/AGENTS.md writing block.
  • references/evaluation.md
    : benchmarks; detector logs; release gates; skill-maintenance work.
仅加载任务所需的参考资料：
  • references/slop-taxonomy.md
    ：全面诊断、红队评审、标记清单。
  • references/voice-and-preservation.md
    ：作者样本、技术散文、法律/医疗/金融声明、语气保留。
  • references/density-and-unsummarizability.md
    ：密集型散文、强化论证、用户提出的「不可概括」标准。
  • references/writing-system.md
    ：基于正向体系生成散文、Orwell的六条规则、被动语态减少、可移植的CLAUDE.md/AGENTS.md写作模块。
  • references/evaluation.md
    ：基准测试、检测工具日志、发布门槛、技能维护工作。

Script routing

脚本调用

Use scripts when the user asks for testing, when local files are available, or when validating a skill change:
bash
node bin/slopbeth.js benchmark
python3 scripts/deslop_lint.py path/to/text.txt --format json
python3 scripts/orwell_lint.py path/to/text.txt --format json
python3 scripts/preservation_check.py original.txt rewrite.txt --format json
python3 scripts/density_report.py original.txt rewrite.txt --format json
Use 
orwell_lint.py
on a single draft to see passive voice, long words, deletable phrases, and jargon; treat per-rule counts as review signals, not a defect ledger. Use 
signature_score.py
, 
cadence_score.py
, 
semantic_drift.py
, 
unsummarizability_check.py
, 
run_benchmark.py
, and 
orwell_benchmark.py
only on before/after corpora that include candidate outputs. Use 
span_annotation_check.py
, 
false_positive_check.py
, and 
competitor_output_score.py
when maintaining the bundled benchmarks.
Load 
references/evaluation.md
for the full benchmark and detector-evidence rules. In this package, use 
scripts/
relative to the installed Slopbeth skill directory.
The scripts report signals. They do not decide whether prose is good enough.
当用户要求测试、本地文件可用或验证技能变更时，使用脚本：
bash
node bin/slopbeth.js benchmark
python3 scripts/deslop_lint.py path/to/text.txt --format json
python3 scripts/orwell_lint.py path/to/text.txt --format json
python3 scripts/preservation_check.py original.txt rewrite.txt --format json
python3 scripts/density_report.py original.txt rewrite.txt --format json
使用
orwell_lint.py
检查单份草稿中的被动语态、长词、冗余短语和行话；将每条规则的计数视为评审信号，而非缺陷清单。仅在包含候选输出的前后语料库中使用
signature_score.py
、
cadence_score.py
、
semantic_drift.py
、
unsummarizability_check.py
、
run_benchmark.py
和
orwell_benchmark.py
。在维护捆绑基准测试时，使用
span_annotation_check.py
、
false_positive_check.py
和
competitor_output_score.py
。
加载
references/evaluation.md
获取完整的基准测试和检测工具证据规则。在此包中，相对于已安装的Slopbeth技能目录使用
scripts/
。
脚本仅报告信号，不决定文本是否足够优质。

Hard rules

硬性规则

  • Never claim text is permanently undetectable, guaranteed human, or safe against all AI detectors.
  • Reject detector tricks that make the writing less true, less specific, or less like the author.
  • Keep vague copy evidence-bound. If concreteness requires missing source material, ask for it or label the example as a placeholder.
  • Do not launder vague outcomes into polished claims. If the source gives only abstract benefits, name the missing mechanism, owner, metric, changed step, or evidence instead of restating the benefit as true.
  • Leave support, recruiting, incident, product, strategy, and education copy without invented owners; dates; failure modes; workflow steps; product surfaces; company names; metrics; or obligations.
  • In support copy, do not add process promises such as "we will review," "we will follow up," or "we will resolve" unless the source says that team action is available. Ask for the required next input and preserve promise boundaries.
  • In policy and incident copy, do not add quality labels such as "auditable," "secure," "resilient," or "controlled" unless the source states that property directly. Keep the rule or incident boundary concrete.
  • Preserve qualifiers that carry scope; uncertainty; causality; risk; or legal/technical meaning.
  • Avoid replacing AI slop with a new formula: clipped aphorisms; tidy triads; forced contrast; dramatic fragments; or generic consultant voice.
  • Over-editing already strong human text is a failure. A light edit or "leave this alone" can be the correct output.
  • Instructions about the writing are not the writing. If the source says what the piece should or should not do, do it; do not print it. "Leave this alone" never means "hand the brief back".
  • Mark exact spans when reviewing long or risky text: bad span; label; reason; preserved span; reason. If the exact span cannot be pointed to, treat the critique as too vague.
  • Check cadence before finalizing medium or long rewrites. Repeated sentence lengths, polished transition stacks, and repeated openers can be slop even when the words are not banned.
  • Avoid em dashes, emojis, title-case hype headings, and decorative bold unless the user's sample clearly uses them and the medium calls for them.
  • Keep the skill's internal checklist shape out of final prose. User-facing rewrites should not default to title-case sections; labeled vertical lists; exhaustive caveat blocks; or polished three-part scaffolds.
  • For detector-facing work, record structured rows with tool name; URL; date; text hash; raw result or screenshot path; result class; and limitation.
  • 绝不要声称文本可永久规避检测、保证是人类创作或能安全通过所有AI检测工具。
  • 拒绝会降低文本真实性、具体性或偏离作者风格的检测规避技巧。
  • 模糊文本需绑定证据。如果具体化需要缺失的源材料，请索要或标记示例为占位符。
  • 不要将模糊的成果粉饰为精美的声明。如果源文本仅给出抽象益处，请指出缺失的机制、负责人、指标、变更步骤或证据，而非将益处重述为事实。
  • 对于支持、招聘、事件、产品、战略和教育类文本，不得编造负责人、日期、故障模式、工作流程步骤、产品界面、公司名称、指标或义务。
  • 在支持类文本中，除非源文本明确说明团队可采取相关行动，否则不要添加诸如「我们将审核」「我们将跟进」或「我们将解决」等流程承诺。索要所需的后续输入并保留承诺边界。
  • 在政策和事件类文本中，除非源文本直接说明相关属性，否则不要添加「可审计」「安全」「resilient」「受控」等质量标签。保持规则或事件边界具体化。
  • 保留承载范围、不确定性、因果关系、风险或法律/技术含义的限定词。
  • 避免用新的公式替代AI冗余内容：简短格言、整齐的三元组、生硬的对比、戏剧化片段或泛化的顾问语气。
  • 过度编辑原本优质的人类文本是失败的做法。轻度编辑或「保持原样」可能是正确的输出结果。
  • 关于写作的指令不属于写作内容本身。如果源文本说明作品应该或不应该做什么，请按要求执行，但不要输出这些指令。「保持原样」绝不意味着「返回需求说明」。
  • 审阅长文本或高风险文本时标记精确范围：问题范围、标签、原因、保留范围、原因。如果无法指出精确范围，则视为点评过于模糊。
  • 在完成中等或长篇重写前检查节奏。重复的句子长度、精心打磨的过渡堆叠和重复的开头即使未使用禁用词汇，也可能属于冗余内容。
  • 除非用户样本明确使用且媒介要求，否则避免使用破折号、表情符号、标题大小写的夸张标题和装饰性粗体。
  • 不要将技能内部的检查清单格式带入最终文本。面向用户的重写不应默认使用标题大小写的章节、带标签的垂直列表、详尽的警告块或精心打磨的三部分框架。
  • 对于面向检测工具的工作，记录结构化条目，包含工具名称、URL、日期、文本哈希、原始结果或截图路径、结果类别以及局限性。