yanghu819

XHS + Hard Thing 4-part 生产流水线技能。统一入口执行抓取、judge、NotebookLM 产出、飞书分发与验收。

Split long books into NotebookLM-ready chunks and batch-generate per-part materials (report, quiz, flashcards, video) with reproducible manifests. Use when you need chapter-level decomposition, controlled source ingestion, and deterministic content packages for education, publishing, or paid knowledge products.

生成“今日 30 条 AI News”（Google News RSS + when:1d），支持研究/工程向筛噪声（过滤荐股/市场报告/泛科普），输出到文件并可推送飞书。

4090 一键跑 delta-embedding benchmark：本机缓存 HF -> rsync 上云 -> uv venv -> 产出 metrics.json（Template vs Content）。

英文 EPUB 翻中双风格（精确保真/短剧精华）实战 SOP：抽样冒烟、全量续跑、抗 400/超时、卡段恢复、最终多版本导出。

在 fengvideo 项目中标准化产出 E01 权威风格视频（左文右像、黑金高亮、MiniMax 配音），支持 25 秒冒烟片与完整主片。

OpenClaw 飞书排障与配置：access not configured（pairing）、open_id cross app、获取 chat_id、发送测试消息/附件、群聊 allowlist。

Reusable CDP-style Xiaohongshu operations patterns for archive/publish/search workflows, with runtime safety guards (single-instance lock) and smoke checks. Use when users ask to stabilize XHS automation, add ops-grade CLI flows, or benchmark against repos like white0dew/XiaohongshuSkills.

Generate and refine high-hook Xiaohongshu conversion copy with red-team/blue-team iteration. Use when users ask for 狠改文案, 去AI味, stronger hooks, or conversion-oriented card copy.

Render Xiaohongshu cards directly with large readable typography and component-based layout (1080x1440). Use when users request 大字、统一风格、避免小图不可读、direct rendering.

Archive a Xiaohongshu user profile into a local folder, including notes, note titles/bodies, images, and comments. Use when users ask to download/backup 全部笔记内容 from a profile URL.

Run a reusable book-swap smoke pipeline for Xiaohongshu cards from outline to evidence assets, markdown, 10-card render, and manifest validation, including optional persona anchor rendering. Use when users ask whether the pipeline can quickly switch books, keep a人物锚点 style, want a new-book dry run, or request 冒烟验证 like 《从0到1》.

Generate and render multiple Xiaohongshu variants, then build preview sheets for side-by-side review. Use when users ask for 多版本预览, A/B/C/D style comparison, or first-card-only comparison.

Refine Xiaohongshu card markdown with local Claude Opus using red/blue/layout/cover passes, then hand off to direct renderer. Use when users ask for 狠改文案、去AI味、提高点击、避免重叠和短句可读.

Convert a Markdown file into Xiaohongshu (Little Red Book) image cards with deterministic layout, preserving markdown structure order and embedded local images. Use when users ask for direct MD-to-XHS card conversion, markdown poster cards, 图文卡片拆分, or preserving original markdown text/format in social image outputs.

小红书笔记素材创作技能。当用户需要创建小红书笔记素材时使用这个技能。技能包含：根据用户的需求和提供的资料，撰写小红书笔记内容（标题+正文），生成图片卡片（封面+正文卡片），以及发布小红书笔记。

从书籍内容生成小红书10张轮播卡（1080x1440），支持 editorial 主题、红蓝文案迭代、动态引用框。Codex/Claude Code 通用。

Universal local-first pipeline for Xiaohongshu content production and backup across book extraction, markdown structuring, 3:4 carousel rendering, publish dry-run, and repository backup. Use when user asks to create, update, or run a reusable XHS skill bundle for Codex and Claude Code.