Borrowing it
Nothing to install: this file belongs to dongbeixiaohuo/writing-agent. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dongbeixiaohuo/writing-agent/main/.claude/skills/style-modeler/SKILL.mdgit clone --depth 1 https://github.com/dongbeixiaohuo/writing-agentWrote 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/dongbeixiaohuo/writing-agent/style-modeler)<a href="https://agentmods.dev/skills/dongbeixiaohuo/writing-agent/style-modeler"><img src="https://agentmods.dev/badge/skills/dongbeixiaohuo/writing-agent/style-modeler/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/dongbeixiaohuo/writing-agent/style-modeler"><img src="https://agentmods.dev/badge/skills/dongbeixiaohuo/writing-agent/style-modeler.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 | $0.00085 | $0.02287 |
| Opus 5.5 | $0.00034 | $0.00915 |
| Sonnet 5.5 | $0.00017 | $0.00457 |
| Haiku 4.5 | $0.00009 | $0.00229 |
Grade A, and why
style-modeler scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.
风格建模专家
只根据用户提供或明确允许获取的样本建模。把作者身份用于样本分组和风格库检索，不把外部履历、既有印象或样本之外的观点写进风格档案。
0. 解析可写风格根目录
开始前必须执行：
python "${CLAUDE_SKILL_DIR}/scripts/manage_style_registry.py" resolve-root --workspace-root "."
把命令唯一输出记录为 [风格根目录]，后续档案、证据和登记表全部使用这个根目录：
- 在本仓库开发环境中返回
claude-runtime/styles，它是唯一源；完成后再运行同步脚本更新消费端镜像。 - 在插件安装到普通工作区时返回
.claude/styles，不得写入插件缓存。 - 禁止跳过解析后直接假定
.claude/styles，否则会造成运行时镜像漂移。
1. 收集样本并确认作者锚点
- 接收本地文档或 URL。URL 先调用
web-article-extractor，将正文保存到docs/YYYY-MM-DD-文章标题.md。 - 批量 URL 按“导航 → 等待 → 提取 → 保存”的顺序串行处理；不要在同一页面上下文并发导航。
- 从每篇样本的作者、公众号和来源 URL 提取作者锚点。
- 在合并文本前核对全部锚点：
- 作者锚点一致：继续建模。
- 作者锚点不一致：按作者分组；无法可靠分组时停下，请用户确认是否真的要建立“混合风格”。
- 作者未知：保留来源标识，不猜测作者。
- 检索
[风格根目录]/：新作者新建档案；已存在的作者读取旧档案和旧证据账本，执行增量验证。
2. 先建立证据账本
在任何风格归纳前，将证据保存到 [风格根目录]/_evidence/[风格名]_evidence.md。每条记录包含：
| 字段 | 要求 |
|---|---|
| 观察现象 | 样本中具体出现的写法 |
| 原文证据 | 支撑判断的必要短摘录 |
| 出现位置 | 样本标题、编号或 URL |
| 功能判断 | 该写法在上下文中的作用 |
| 可复用规则 | 如何在新主题中复现 |
| 适用边界 | 什么时候适用或不适用 |
硬规则：
- 稳定特征至少要有 2 条证据，且来自至少 2 篇独立样本。
- 单篇样本只能产出“候选特征”，不能宣称是稳定风格。
- 不把样本主题等同于作者风格。
- 每条画像、判断和使用规则都必须能回指证据账本。
- 原文摘录只保留证明结论所需的最短片段，不把长段原文当模板堆入风格文件。
证据分级、增量更新和先验隔离规则见 style-core-and-validation.md。
3. 运行量化指纹
先运行脚本，再写句长、段落、问句和标点结论：
python "${CLAUDE_SKILL_DIR}/scripts/style_fingerprint.py" docs/[样本1].md docs/[样本2].md
将结果写入风格档案的 00b. 量化指纹，并注明样本数。节奏描述必须与实测一致；句首词至少出现 3 次才可列入词汇指纹。
4. 提炼风格，而不是通用好文章公式
按 15-dimensions.md 分析 15 个维度，并额外回答：
- 作者面对新题目首先观察什么？
- 作者会自动排除哪些角度、素材和建议？
- 从现象到判断的论证发动机是什么？
- 素材选择偏好和拒绝项是什么？
- 情绪曲线和读者关系是什么？
- 换一个新主题时，最小可复现配方是什么？
- 多篇样本反复出现的实质判断和价值排序是什么？
对每条核心规则执行两道检查：
- 通用基线检查：任何优秀写作者本来就会做的规则，移到“通用基线”，不算作者指纹。
- 库内区分检查：与两个以上既有风格重复的规则降级，并写出与最相近风格的“区分开关”。
5. 做可证伪的仿写验证
5.1 陌生主题验证
- 选择样本未覆盖的新主题，写 300–500 字短文。
- 从切入角度、素材选择、论证推进、句式节奏、情绪与读者关系五项评分，每项 0–2 分。
- 每个 2 分必须标注对应的证据账本条目；无法溯源按 0 分计。
- 总分低于 8 分，返回风格内核修正。
5.2 多组独立盲测
- 准备至少 3 组功能相近的“原文短段 + 仿写短段”，每组随机打乱 A/B。
- 将每组分别交给不接触风格文件和作者身份的独立评审，只提供两段纯文本。
- 允许评审回答“无法判断”；每次必须给出判断、置信度和有证据的破绽，不要求凑满固定数量。
- 汇总而不是单次判定：
- 至少 2 组判断错误或无法判断，且没有高置信度的“判断方式”破绽：通过。
- 至少 2 组正确识别，或出现高置信度的判断方式破绽：回炉修正风格内核。
- 用量化脚本复测仿写段；关键节奏指标偏离样本基线超过 30% 时不得通过。
What ships with it
7 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.
- references/15-dimensions.md 9.6 KB
- references/style-core-and-validation.md 7.1 KB
- references/style-output-template.md 3.1 KB
- scripts/manage_style_registry.py 6.4 KB runs code
- scripts/normalize_style_frontmatter.py 6.3 KB runs code
- scripts/style_fingerprint.py 6.2 KB runs code
- test-prompts.json 1.9 KB
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 · 161 lines · 85 tokens per session scan A ae8ea86bec91
style-modeler is a skill published in the GitHub repository dongbeixiaohuo/writing-agent (442 stars, last pushed 5d ago), licensed MIT. It adds 85 tokens to every session and 2,287 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
webnovel-init
A guided setup process for a web-novel project that collects creative requirements and creates the project’s initial structure and rules.
webnovel-write
A workflow for producing publishable web-novel chapters, from understanding the context through drafting, review, polishing, submission, and backup.
webnovel-plan
A planning tool for turning a web-novel outline into volume plans, timelines, and chapter outlines. It also writes new story settings back into the existing setting collection.
webnovel-review
A review agent for assessing the quality of fiction chapters. It produces a review report and writes review measurements back to the project.
webnovel-query
A Chinese-language skill for querying a web novel project's setting and story information, including characters, power systems, factions, and foreshadowing.
webnovel-learn
A project skill for learning from a web novel-writing project through the `/webnovel-learn` command. The provided details do not explain what learning tasks or outputs it performs.