视觉画布设计
canvas-design
基于设计原则创作静态视觉作品,用于海报与品牌图形。
content【文档与办公】Turn any content into 3-4 finished LinkedIn posts ready to publish. Use this skill whenever the user shares something and wants LinkedIn posts from it — a PDF, an email, a newsletter, a URL, pasted text, a report, or anything else. Trigger on phrases like "turn this into posts", "what can I post about this", "LinkedIn content from this", "post ideas"…
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Turn any content the user shares into 3-4 finished LinkedIn posts, ready to publish.
Setup note: This skill pairs with the
postskill, which carries the user's voice profile. Make sure both are installed. References below assume both skills live under your skills directory (e.g.~/.claude/skills/post/SKILL.mdand~/.claude/skills/factcheck/SKILL.md) — adjust the paths if yours are different.
Identify what the user has provided and handle accordingly:
Pasted text or notes — use directly from the conversation.
PDF or uploaded file — read the pdf-reading skill (if available) for extraction instructions before proceeding.
URL or web page — use web_fetch to retrieve the full page content.
Email — search Gmail using relevant terms from the user's description. Fetch the full thread.
Newsletter — if the user has a recurring newsletter they pull from, search Gmail for the sender address and fetch the most recent edition. Extract plain text body; ignore tracking pixels, unsubscribe links, and HTML artifacts.
Nothing provided — ask the user what content they'd like to repurpose, or default to whatever recurring source you've been set up to use.
Read through the content and find the strongest raw material across these four angles. Not every piece will have all four equally — pick the best 3-4 total, don't force weak ones.
Critical rule: use the user's exact words wherever possible if they wrote the source (newsletter, email, notes). Lift sentences, transitions, and analogies directly — don't paraphrase their phrasing into something blander. If the source is third-party (a PDF, article, URL), extract the core ideas and write in the user's voice using the post skill.
For each post:
post skill (read its SKILL.md before drafting)Write full, un-hatcheted drafts first. Paragraph form. Do not pre-trim.
Before presenting, identify any posts containing statistics, named figures, or specific claims attributed to organizations or surveys. Run the factcheck skill on those posts (standard tier).
Read the factcheck SKILL.md and verify all checkable claims via web search.
Apply any corrections directly into the drafts before presenting. Note what was fixed and why — one line per fix, after the posts.
Present all posts in full, ready to copy-paste. Label each with its angle type.
After the posts, include a single short note on any factcheck fixes applied.
Ask the user which (if any) they want to adjust, shorten, or take in a different direction.
hatchet automatically — only shorten if asked