anbanwriter-claudecode

Use when user mentions "初始化", "anban-setup", "第一次使用", "API Key", "密钥", or when MCP tools fail with auth/connection errors suggesting missing ANBAN_API_KEY.

Creates and manages WeChat news article drafts (图文草稿) with HTML formatting. Use when creating or managing WeChat news article drafts. Also use when user mentions '发草稿', '发布文章', '创建草稿', 'publish draft', '草稿箱', or when the article pipeline reaches the draft publishing step.

Configures anban-creator project settings including writer style (写作风格), theme (排版主题), image provider (图片生成服务), and API keys. Use when user wants to initialize, view, or modify anban-creator configuration. Also use when user mentions '配置项目', '设置风格', '修改配置', 'config', '项目设置', '查看配置', '写作风格', '排版主题', or when wanting to check or change project-level settings.

Writes WeChat articles with style guidance, removes AI traces (去AI痕), converts Markdown to WeChat HTML, and checks content compliance (合规检查). Use when writing articles, removing AI traces, converting Markdown to WeChat HTML, or checking content compliance. Also use when user mentions '写文章', '去痕', '去AI味', 'HTML转换', '违禁词检查', '合规检查', 'humanize', or when any step in the article pipeline calls for writing, decontaminating, or converting article content.

电商产品图分析——从多张产品图抽取电商转化相关属性，构建锁定规格「产品档案 Product Bible」，作为产品跨图一致性与卖点提炼的共同数据底座。当电商出图流程需要理解产品、构建产品档案时使用。

电商出图全自动创作。用户提到"电商出图"、"电商素材"、"商品图"、"产品图"、"主图"、"详情页"、"商详"、"SKU图"、"电商封面"、"电商设计"、"ecommerce"时使用此 skill。

Use when coloring line art images, batch coloring multiple images, preserving visual consistency across characters, or when user mentions "线稿上色", "上色", "填色", "coloring", "color consistency", "批量上色", "角色上色", "给线稿上色".

Use when generating multiple pose/expression variants from a single portrait photo while keeping the person's identity consistent, or when user mentions "人像姿态", "人像一致性", "同一人物不同表情", "封面人物表情包", "人像变体", "portrait consistency", "pose variants", "手势变化", "表情封面". Triggers whenever a user provides one portrait photo and asks for multiple cover-ready variants of the same person — even if they don't explicitly say "一致性". Generates 1-6 vertical 9:16 cover-ready portraits from one reference photo, locked to the same person's face/features/hair/style.

Use when replicating viral short-video covers, generating a short-video cover from a reference cover image, or when user mentions "短视频封面", "爆款封面", "封面复刻", "复刻封面", "cover replication", "B站封面", "抖音封面", "视频号封面", "小红书视频封面". Triggers whenever a user provides a reference cover image and asks for a new short-video cover based on it — even if they don't explicitly say "复刻". Covers the 9:16 vertical cover replication workflow: analyze reference cover's visual logic → migrate to user's new title → generate cover prompt → quality optimization.

Use when editing source footage into finished videos by conversation: 多素材成片剪辑, talking-head cleanup, 去口癖, retake selection, subtitles, color grade, animation overlays, social-ready exports, or transcript-backed video editing with OpenAI-compatible FunASR MCP tools.

Use when generating commercial short videos with 即梦, Dreamina, Seedance, Seedance 视频生成, 种草视频, 带货视频, 获客视频, 推广视频, 图生视频, 多模态视频, or when a user provides image/audio/video references and asks for a stable marketing video.

Use when video-use needs agent-authored HTML/CSS/JavaScript motion graphics, kinetic typography, UI motion, product callouts, or animated cards rendered with HyperFrames for a video overlay.

Use when video-use needs precise explanatory animation, diagrams, formulas, graph motion, process visualization, or educational overlays rendered with Manim.

Use when video-use needs deterministic lightweight overlays made from Pillow/PIL images, PNG sequences, simple text cards, counters, progress bars, or reveal frames.

Use when video-use needs React/Remotion-based parameterized motion graphics, reusable video templates, data-driven animated cards, or component-based overlays.

微信公众号图文文章全自动创作。用户提到"写文章"、"写一篇"、"发文章"、"公众号文章"、"推文"时使用此 skill。

微信公众号封面图（公众号头图）专用设计——硬编码官方比例 900×383（2.35:1），中心安全区构图保证转发卡 1:1 主体完整，受控文字策略（按真实场景决定是否带短文字），从文章核心隐喻推导视觉概念，生成后 vision 6 维评分卡把关。用户提到「封面」「公众号封面」「头图」「cover」「封面设计」「封面图」时，或 article 流水线到达封面生成步骤时使用。

微信公众号爆款互动引擎——为文章注入完读率/转发率/收藏率/评论率四大互动驱动力。覆盖选题社交货币判定、标题CTR（好奇心缺口+情绪+数字）、黄金三秒开头钩子、情绪弧与金句密度、完读率节奏、转发/收藏/评论诱因、发布前爆款审计。当公众号文章创作流进入选题、写作、标题优化或发布验收任一环节时务必使用此 skill；凡是用户想提升公众号文章的曝光、流量、打开率、点赞、评论、转发、收藏、涨粉的，都必须用此 skill。不要因为"只是写一篇文章"就跳过——爆款不是写完才考虑的，而是从选题、标题、开头、结构每一步都要注入的。

Manages images for WeChat article (公众号图文) content including template selection, visual rhythm planning, cover generation (封面), content illustration (配图) with vision verification, compression, and CDN upload. Use when generating or processing images for WeChat articles. Use when user mentions '封面', '配图', '插图', '视觉设计', '图片上传', 'generate cover', 'visual rhythm', 'template', or when the article pipeline calls for image planning, generation, verification, or uploading.

种草笔记图文全自动创作。用户提到"种草笔记"、"seednote"、"种草"、"复刻"、"仿写"、"改写笔记"、"爆款改写"、"克隆"、"clone"时使用此 skill。

Generates cover (封面), content pages (内容图), and tail pages (尾图) for Seednote (种草笔记) posts with 3:4 ratio design norms. Use when creating seednote visual content including covers, content pages, and tail pages. Also use when user mentions '种草笔记图片', '封面生成', '内容图', '尾图', '图片规划', or when the seednote pipeline calls for image generation.

Writes and optimizes Seednote (种草笔记) content including titles, body text, hashtags, and viral note rewrites. Use when writing Seednote content, optimizing seednote titles, or rewriting viral notes. Also use when user mentions '写种草笔记', '笔记文案', '种草文案', '标题优化', '改写笔记', '爆款改写', '复刻文案', '合规检查', or when the seednote pipeline calls for content creation or text optimization.

操控剪映/CapCut 草稿文件——创建、读取、修改、删除剪映项目草稿。用户提到"剪映"、"剪映草稿"、"剪映项目"、"CapCut draft"、"JianYing draft"、"导出剪映"、"创建剪映草稿"、"修改剪映草稿"时使用此 skill。当用户需要批量操作剪映项目、自动化视频编辑工作流、或程序化管理剪映草稿时也应触发。

Use when working on live video slicing, 直播切片, 剪直播, 智能切片, 听悟 transcription, or live video slicing workflows that turn long livestream videos into transcript-backed short-video clips.

去除文本中的 AI 生成痕迹（去 AI 味），使文章更自然、更像人类书写。基于维基百科《Signs of AI writing》指南（WikiProject AI Cleanup 维护），检测并修正 33 类 AI 写作模式。当写作流程需要去除 AI 痕迹，或用户提到「去 AI 味」「去痕」「humanize」「润色得像真人写的」时使用。由 content-writing / seednote-writing / ecommerce-copywriting 在写作流程中调用，对中文正文就地改写。

电商转化文案——从产品档案提炼 FABE 核心卖点，生成主图5张结构文案、详情页黄金结构文案、分享文案。受众=买家，目标=点击率→转化率→客单价、降退货。当电商出图流程需要卖点提炼或各模块文案时使用。

电商平台规范与合规知识库——各平台尺寸/主图规则/详情结构 + 《广告法》极限词与平台电商违禁词扫描。当电商出图流程需要注入平台尺寸规范或做合规检查时使用。

电商视觉设计与生成——商业转化导向的视觉工艺，按已选模块规划并生成主图/详情/封面/分享/SKU，保证产品跨图一致。受众=买家，目标=点击→转化→降退货。当电商出图流程需要图片规划或生成时使用。

Analyzes Seednote (种草笔记) topics, trending notes (热门笔记), and scores engagement potential (互动率评分). Use when analyzing Seednote topics, scoring engagement, researching trending seednote content, or fetching source note details for replicate mode. Also use when user mentions '种草笔记选题', '热门笔记', '竞品分析', '笔记分析', or when the seednote pipeline calls for topic discovery or source note fetching.

Use when analyzing or decomposing a Seednote (种草笔记) viral note, extracting reusable templates from a source note, or preparing evidence-backed clone guidance. Also use when the user mentions "拆解爆款", "爆款拆解", "分析爆款笔记", "提取爆款模板", "source note analysis", or when the seednote replicate pipeline needs source-note analysis.

Researches WeChat topics (选题研究), scores engagement potential, and generates content outlines (大纲). Use when researching topics, scoring engagement potential, or generating content outlines. Also use when user mentions '选题', '话题研究', '大纲', '选题分析', '话题评分', or when any content pipeline calls for topic discovery, scoring, or outline generation.

Optimizes titles (标题优化), keywords (关键词), and article digests (摘要) for WeChat search ranking. Use when optimizing titles, keywords, or writing article digests for search ranking. Also use when user mentions 'SEO', '优化标题', '关键词优化', '摘要', '搜索排名', or when the article pipeline reaches the SEO step.