agentleFS
Gingiris-1031Gingiris-1031/gingiris-skills/skills/gr-brand-voice-calibration/SKILL.md

gr-brand-voice-calibration

Skill89 starsChanged 32 days agoMITView on GitHub

Use this skill to calibrate an author''s or brand''s writing voice from original samples and edits,

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use authorized local originals before generic web research.

    line 24
  2. Distinctive:

    Use evidence worksheet to distinguish observation, hypothesis and a confirmed preference.

    line 34
  3. Distinctive:

    Do not invent mandatory time-place scenes, first-person openings, punctuation counts or stats tables.

    line 42
  4. Distinctive:

    When deriving rules from high/low-performing posts, use performance hypotheses.

    line 70

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Brand Voice Calibration / 文风校准

Learn how this person actually expresses a thought, not a generic “high-performing content” formula. Voice does not authorize publishing or supply facts. Follow the user's language and platform request.

学习作者如何真实表达,不把爆款套路当人格。文风不构成发布授权,也不能提供文章事实。跟随用户语言与平台要求。

1. Establish source priority / 确认来源层级

Current instructions and the author's own corrections → original speech/writing in context → explicitly approved guide → older guides → edited/AI drafts and third-party suggestions. Respect a different priority explicitly chosen by the author. Keep repost commentary separate from quoted authors; identify speakers before extracting transcript samples.

当前要求与亲手修改 → 有完整语境的本人原话/原文 → 已认可指南 → 历史指南 → 编辑/AI 草稿与第三方建议。用户可明确调整优先级。转发评论与被引用作者分开,逐字稿先确认说话人。提示词、写作指令和幻灯片风格要求不自动是作者原声。

Use authorized local originals before generic web research. Record duplicate or edited variants; an exported row is not necessarily one original post. Do not download or publish the entire personal archive merely to build a guide.

优先使用获准的原始资料,记录重复与改写版本;导出一条不等于一篇原创。没有必要就不下载图片,也不公开整个私人素材库。

2. Build a contextual sample / 建立有语境的样本

Sample across time, genre, audience and emotional register as available. Keep original text, surrounding context, date, author/role and editing status. Start with a few representative examples and a few disliked examples if supplied; do not demand a fixed quota. Label limited samples as provisional. A user's “most liked/disliked” means their judgment unless actual audience metrics were supplied.

按时间、体裁、受众和情绪跨度抽样,保留原文、上下文、日期、说话人和编辑状态。可从几篇认可/不认可样本开始,不强求数量;样本不足则暂定。本人喜欢不等于数据表现最佳,互动高也不证明最像本人。

Use evidence worksheet to distinguish observation, hypothesis and a confirmed preference.

3. Infer voice with counterexamples / 提炼并找反例

Observe entry points, sentence/paragraph rhythm, vocabulary, judgment/uncertainty, emotional intensity, asides, relationships with readers, endings and omissions. For each proposed rule retain an example, counterexample or scope limit. Distinguish voice from topic, platform limits and publishing tactics.

观察开头、句段节奏、用词、判断与不确定、情绪、旁白、与读者关系、结尾和留白。每条规则保留例子及反例/适用边界;声音、题材、平台限制和分发策略分开。

Do not invent mandatory time-place scenes, first-person openings, punctuation counts or stats tables. Repetition, enthusiasm and unfinished thoughts can be authentic. Do not flatten everything into short sharp consultant prose or add an uplifting conclusion the author never made.

不硬造时间地点、第一人称开头、标点配额和数字表。重复、热烈和没想完也可能是真实表达;不要一律改成短而锋利的顾问口吻,更不擅加升华。

4. Draft and test / 试写与校验

Produce a concise guide with stable traits, contextual variations, don't-do examples, evidence anchors and unresolved preferences. Select a source-backed topic; write a short test in the requested genre. Keep some original samples out of guide construction to check whether the inferred rules generalize. Compare a minimally edited version with the adapted version when useful.

交付简明指南:稳定特征、场景变化、反例、来源锚点和待确认偏好。选择有事实来源的主题试写;留一部分原文不参与规则提炼,用来检查规则是否过拟合。必要时对比轻编辑稿与改写稿。

Check meaning, factual fidelity and preserved intent before style. Ask for focused feedback on specific passages; revise rules only when supported by edits or evidence. An agent's self-rating is not author approval. If no author feedback, label the guide provisional and provide a usable draft anyway.

先查意思、事实与原意,再看风格。对具体段落征求反馈并据此修正规则;AI 自评分不等于作者认可。没有反馈也交付可用稿,但指南标待确认。

5. Cross-language transfer / 中英文迁移

Preserve stance, specificity, warmth, uncertainty and degree of endorsement. Rebuild natural rhythm in the target language instead of copying Chinese punctuation or translating every aside. Keep an invariant meaning/source checklist. If no native samples exist, describe the result as an adaptation, not a verified native-language replica.

保留立场、具体程度、温度、不确定与推荐力度,重建目标语言自然节奏,不机械翻译括号和破折号。逐项核对意义和事实不变。没有目标语言原声样本就标为适配稿,不宣称已验证“完全像本人”。

Deliver / 交付

Source inventory + compact voice guide + source-backed trial + differences/feedback log. Only publish the general method or explicitly authorized guide; private samples and sensitive details stay in the user's approved storage. Downstream WeChat/blog/social workflows consume the guide but must independently verify facts, format and publication authorization.

交付来源索引、简明文风指南、有来源试稿及修改记录。公开 Skill 只含通用方法,私人样本不随包分发;下游公众号/博客/社媒流程仍需独立核实事实、排版和发布授权。

Performance samples / 表现样本

When deriving rules from high/low-performing posts, use performance hypotheses. 区分本人偏好与分发表现;语言、主题与时效同时变化时,不归因给单一文风因素。

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.