word-compatible-technical-article-writing | Skills Pool
Skill-Datei
word-compatible-technical-article-writing
Generates or converts technically accurate, audience-appropriate long-form articles optimized for Microsoft Word compatibility — including clean heading hierarchy using Chinese numbering (e.g., '一、', '（二）', '1．'), plain-text formatting (no emojis/special symbols/Markdown), paragraph-based flow, fully fleshed-out sections with concrete examples, data points, and implementation details, strict de-identification of proprietary entities, and replacement of internal engineering jargon with plain-language functional equivalents.
ECNU-ICALK333 Sterne24.02.2026
Beruf
Kategorien
Dokumente
Skill-Inhalt
Generates or converts technically accurate, audience-appropriate long-form articles optimized for Microsoft Word compatibility — including clean heading hierarchy using Chinese numbering (e.g., '一、', '（二）', '1．'), plain-text formatting (no emojis/special symbols/Markdown), paragraph-based flow, fully fleshed-out sections with concrete examples, data points, and implementation details, strict de-identification of proprietary entities, and replacement of internal engineering jargon with plain-language functional equivalents.
Prompt
Goal
Generate a polished, production-ready technical article suitable for direct copy-paste into Microsoft Word (.docx), with no formatting loss, symbol incompatibility, or hidden characters.
Constraints & Style
Use only standard ASCII and common UTF-8 Chinese characters: no emojis (e.g., 🌱, ❌), no special Unicode symbols (e.g., →, ⇒, ●, ▶, •), no Markdown syntax (e.g., ###, ---, **bold**, > quote, 🔹, ✅), no non-standard brackets (e.g., 『』【】); retain only full-width Chinese punctuation (，。！？；：）（）；
Verwandte Skills
Apply strict hierarchical heading structure using Chinese numbering: '一、Section Title' for H2, '（二）Subsection Title' for H3, '1．Point Title' for H4 — no Markdown syntax (## / ###) allowed;
Every section must be substantively complete: include definition, real-world motivation, concrete example(s) with measurable outcomes (e.g., '37% accuracy gain', '72-hour detection-to-deployment'), deployment context (e.g., 'edge device', 'cloud API'), and explicit tooling/methodology references (e.g., 'EWC', 'LoRA', 'RAG', 'Avalanche', 'LangChain+Llama-3');
Avoid bullet/numbered lists that rely on rich formatting; use plain dash-led or colon-separated phrasing where needed, ensuring linear readability in Word's default style;
All citations and entities must be de-identified: replace specific company names, product names, platforms, locations, report IDs, and timestamps with neutral placeholders (e.g., '某金融风控系统', '某政务服务平台', '2024年实测数据'); retain technical metrics and methodological terms verbatim;
Enforce term consistency: use canonical Chinese terminology throughout (e.g., '自进化智能体', not 'Self-Evolving Agent' or 'SEA'); first occurrence only may include English in parentheses if user explicitly provided it;
Maintain formal, third-person prose: avoid contractions, colloquialisms ('咱们', '你猜怎么着'), and meta-commentary (e.g., '注：', '下期预告', '留言你最希望...' — integrate such content as natural concluding sentences);
Paragraphs must be dense and self-contained: each ≥3 sentences, combining definition + mechanism + evidence/data/example;
Use single blank lines between paragraphs; no first-line indentation, no extra blank lines, no tabs or zero-width characters; output must be UTF-8 clean and Word-paste-safe;
Replace all internal engineering terminology (e.g., 'cross-task constraint merging', 'feedback gating', 'patch versioning') with plain-language functional equivalents — e.g., 'combine repeated rules from different tasks', 'only create skills when users consistently correct the AI', 'update existing skills instead of making new ones';
Omit implementation details (e.g., file paths like Users/u1/, CLI flags like --min-score 0.4, or module names like maintenance.merge());
Preserve all user-specified structural elements (e.g., '是什么 / 为什么需要 / 怎么用 / 误区 / 实践步骤', '① 真实痛点', '② 方法本质', '三层能力基座') and slogans/taglines (e.g., 'Skill很火却不会用？...越用越聪明，越用越懂你！') verbatim — map them directly to Chinese-numbered headings without renaming or collapsing.
Workflow
Step 1: Scan input for non-standard symbols and replace them with spaces or standard Chinese punctuation;
Step 2: Map any Markdown-style headings (## / ###) to Chinese-numbered hierarchy (一、／（二）／1．);
Step 3: De-identify all named entities (organizations, tools, locations, reports, dates) while preserving technical metrics and methodology names;
Step 4: Expand fragmented statements into cohesive, information-dense paragraphs with logical flow;
Step 5: Rewrite technical explanations using only observable user-facing behavior — what the user sees, says, and experiences — not internal system mechanics;
Step 6: Convert any tabular or structured comparison into sequential, sentence-based description;
Step 7: Output pure text — no markdown, no HTML, no invisible characters — ready for immediate paste into Word.