Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add konglong87/shenbi-maliang --skill writing-dnagit clone --depth 1 https://github.com/konglong87/shenbi-maliangWrote 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/konglong87/shenbi-maliang/writing-dna)<a href="https://agentmods.dev/skills/konglong87/shenbi-maliang/writing-dna"><img src="https://agentmods.dev/badge/skills/konglong87/shenbi-maliang/writing-dna/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/konglong87/shenbi-maliang/writing-dna"><img src="https://agentmods.dev/badge/skills/konglong87/shenbi-maliang/writing-dna.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00124 | $0.05103 |
| Opus 5 | $0.00062 | $0.02551 |
| Sonnet 5 | $0.00025 | $0.01021 |
| Haiku 4.5 | $0.00012 | $0.00510 |
Grade A, and why
writing-dna 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 2d 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 — 348 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🖌️ 神笔马良 · 写作DNA蒸馏
将自己公众号/小红书/X/Twitter 的文章蒸馏成属于你自己的「写作基因包」。 核心路径：语料准备 → 七维并行采集 → 三重验证降噪 → 硬规则/软范式分类 → 分层产物 → 复刻写作 → 成稿收敛
理念来源：借鉴认知蒸馏方法论，将隐性的个人写作风格， 转化为 AI 可直接执行的显性规则，本质是人文本范式的萃取与固化。
v2 关键变化：产物从单一档案改为分层产物；写作前增加硬性必读流程； 成稿后增加去 AI 味收敛工序。蒸馏解决「像不像你」，收敛解决「像不像人写的」。
技能触发场景
- 用户提供文章样本，说"蒸馏文章"或"分析我的写作风格"
- 用户说"按我的风格写一篇文章"
- 用户新增文章，说"更新我的DNA"（增量蒸馏）
- 用户查看当前DNA档案
- 用户说"帮我找出我写作的盲区/反模式"
- 用户说"这篇太AI了"或"去掉AI味"（成稿收敛）
阶段零：语料准备与元数据标注
蒸馏质量的上限由语料决定。样本不足或没有元数据，后面所有步骤都是空中楼阁。
语料要求
- 数量：至少 20 篇完整文章（低于 10 篇只能得到粗略印象，不可当真）
- 格式：
.md或.txt - 覆盖：尽量包含不同时期、不同体裁、不同主题的文章
目录结构
用脚本一键生成，或复制 templates/author-corpus/：
corpus/<作者名>/
├── raw/ 原始文章语料
├── _meta/ 每篇一份结构化标注
├── 语言基因.md L1 表层语言
├── 结构骨架.md L2 文章结构
├── 认知素材.md L3-L5 选题·素材·认知
├── 视觉排版.md L6 排版与配图（图文类账号必做）
├── 反模式.md L7 盲区与禁忌（本项目独有）
└── 写作DNA.md 整合文档，≤4000 字
raw/ 命名规范：YYYY-MM-DD 体裁 文章标题-来源.md
元数据标注（不可跳过）
每篇文章在 _meta/ 建一份标注，schema 见 templates/author-corpus/_meta/example.schema.json：
{
"title": "", "date": "YYYY-MM-DD", "author": "", "column": "",
"article_type": "访谈|深度分析|短评|观察|综述|教程|故事",
"topic_tags": [], "hook_type": "问题式|场景式|数据式|观点式|悬念式|故事式",
"structure_pattern": "总-分-总|时间线|对比式|Q&A|层层递进|问题-分析-方案",
"source_types": [], "word_count": 0, "notable": ""
}
article_type和topic_tags是阶段五「选 5 篇原文校准」的检索键。 没有元数据，那 5 篇就选不准，语感校准就无从谈起。
阶段一：文本预处理
- 格式清洗：去除多余空行、特殊符号、排版标记
- 话题语义切片：不按段落机械切分，按话题单元划分（一个完整论点或叙事片段为一个切片）
- 重复内容识别：标记在多篇文章中反复出现的段落（固定套路 or 口头禅），这类重复是 DNA 的重要信号
- 图片内容纳入：图文类账号必须抽样 5-10 篇逐张查看图片内容。截图、对话记录、数据表格里的文字是论证链的一部分，不看图会漏掉承重结构
阶段二：七路并行特征采集（7-Collector）
读取 references/writing-dna-framework.md 获取每个维度的详细分析指南。
七个维度独立并行分析，各自提取碎片化特征，最终汇入「原始特征池」：
| # | 采集器 | 萃取内容 |
|---|---|---|
| 1 | 表达范式 | 词频、句式习惯、标志性词汇、口头禅、修辞手法、中英混用 |
| 2 | 思维逻辑 | 开头钩子、论证套路、推理路径、转折方式、结尾收束 |
| 3 | 知识与素材 | 高频话题领域、类比素材库、素材来源策略、权威对象选取、数据使用方式 |
| 4 | 情感决策 | 与读者的关系定位、情绪调性频谱、共情策略、价值观优先级 |
| 5 | 选题视角 | 切入时机与角度、标题命名规律、话题优先级、明确不写什么 |
| 6 | 节奏控制 | 句长分布、段落节奏、短句爆破节点、留白风格、标点习惯 |
| 7 | 反模式 ⚠️ | 思维盲区、逻辑漏洞、表达短板、禁忌话题、负面口头禅 |
What ships with it
21 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.
- config.yaml 5.2 KB
- references/ai-tone-rules.md 23 KB
- references/dna-template.md 6.2 KB
- references/human-voice-rules.md 7.1 KB
- references/writing-dna-framework.md 8.6 KB
- scripts/distill_writing_dna.py 49 KB runs code
- templates/author-corpus/_meta/example.schema.json 621 B
- templates/author-corpus/examples/README.md 1.2 KB
- templates/author-corpus/examples/口语闲聊风/写作DNA.md 4.3 KB
- templates/author-corpus/examples/口语闲聊风/反模式.md 2.5 KB
- templates/author-corpus/examples/口语闲聊风/结构骨架.md 3.9 KB
- templates/author-corpus/examples/口语闲聊风/认知素材.md 3.5 KB
- templates/author-corpus/examples/口语闲聊风/语言基因.md 3.5 KB
- templates/author-corpus/raw/.gitkeep 157 B
- templates/author-corpus/README.md 1.6 KB
- templates/author-corpus/写作DNA.md 1.4 KB
- templates/author-corpus/反模式.md 1.5 KB
- templates/author-corpus/视觉排版.md 1.9 KB
- templates/author-corpus/结构骨架.md 1.8 KB
- templates/author-corpus/认知素材.md 2.4 KB
- templates/author-corpus/语言基因.md 2.0 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.
- 2d ago Changed · +187 lines · +50 tokens per session 3242cea1152e
- 12d ago First seen · 161 lines · 74 tokens per session scan A 4c0397df4288
writing-dna is a skill published in the GitHub repository konglong87/shenbi-maliang (2 stars, last pushed 3d ago), licensed MIT. It adds 124 tokens to every session and 5,103 once invoked, about $0.0006 per session on Opus 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 2026-08-31.
Other skills, from other repositories
tech-design-generation
Generates professional Technical Design Documents following industry-standard engineering practices. This skill activates when the user asks to write a technical design, create an architecture design, draft a design doc, write a technical proposal, do system design, create a technical design document, write an RFC…
test-cases-generation
Generates structured test case sets with multi-dimensional coverage from project code analysis or specification documents. This skill activates when the user asks to write test cases, generate tests, supplement tests, create test coverage, or improve test completeness. It auto-scans the project to extract testable…
idea
Use when exploring a new idea before writing formal specifications. Guides users through iterative discovery, research, and critical analysis to crystallize vague ideas into validated requirement drafts. Includes anti-pseudo-requirement checks, competitive research, demand evidence gathering, and "What if we don't…
test-plan-generation
Generates professional Test Plan and Test Cases documents following IEEE 829 test documentation standards, ISTQB testing methodologies, and Google Testing Blog best practices. This skill activates when the user asks to write a test plan, create test cases, draft a test plan document, create a test cases document…
prd-generation
Generates professional Product Requirements Documents (PRDs) following industry-standard methodologies. This skill activates when the user asks to write PRD, create a product requirements document, draft product requirements, do product planning, do feature planning, write a product spec, or create a product document.…
propagate
Walk an upstream documentation change downstream — when a PRD, SRS, tech-design, or feature spec is edited, find every dependent document, identify which sections are now stale, and interactively update them to keep the entire doc chain consistent. Use after editing any spec-forge document, or run periodically with…