anti-ai-writing
Detect and eliminate signs of AI-generated writing. Use when producing any user-facing text — blog posts, landing pages, docs, emails, social media, README, marketing copy. Enforces human-quality prose by flagging Wikipedia's documented AI writing tells.
- License 许可
- GPL-3.0
- 最近更新
- 2026-04-24
- Stars
- ★ 1
- Forks
- 3
01 前言 Front Matter
- name
- anti-ai-writing
- next
- []
- group
- comms
- workflows
- [ "all" ]
- always-load
- true
- description
- Detect and eliminate signs of AI-generated writing. Use when producing any user-facing text — blog posts, landing pages, docs, emails, social media, README, marketing copy. Enforces human-quality prose by flagging Wikipedia's documented AI writing tells.
- prerequisites
- []
- user-invocable
- true
03 SKILL.md 正文
Anti-AI Writing
Every piece of text this harness produces must read like a human wrote it. AI-generated writing has documented tells. This skill flags and fixes them.
Source: Wikipedia's "Signs of AI writing" — the most comprehensive public catalog of AI writing patterns.
Banned Words and Phrases
Never use these. They are the most documented AI tells:
Promotional fluff: vibrant, nestled, rich heritage, groundbreaking, showcasing, diverse array, pivotal moment, stands as, testament, marks a shift, broader trends, cutting-edge, revolutionizing, game-changing, innovative
Vague intensifiers: delve, intricate, tapestry, foster, enhance, leverage, utilize, multifaceted, comprehensive, robust, seamless
False attribution: industry reports suggest, experts argue, researchers note, studies show (without citing specific studies)
Filler transitions: additionally, furthermore, moreover, it's worth noting, it's important to note, in today's landscape, in the realm of
Outline conclusions: "Despite its [positive traits], [subject] faces challenges..." — never use this formula
Structural Tells to Avoid
- Rule of Three abuse — stop listing three adjectives or three phrases for artificial comprehensiveness. Use one strong word.
- Negative parallelisms — "Not just X, but also Y" sounds like a press release. Say what it IS.
- Elegant variation — don't replace a word with a synonym just to avoid repetition. Repeat the word. "The API" is clearer than "the interface" the second time.
- Em dash overuse — use commas or parentheses. Em dashes in every paragraph = AI tell.
- Excessive boldface — bold is for emphasis, not decoration. One bold per paragraph max.
- Title Case headings — use sentence case: "How to deploy" not "How To Deploy"
- Inline-header bullet lists — not everything needs to be a bold-header-colon-description bullet. Write prose.
- Section summaries — never summarize what you just said at the end of a section.
Tone Rules
- Never sound promotional. You're explaining, not selling. "This tool syncs your database" not "This powerful tool revolutionizes database synchronization."
- Never inflate significance. Small features are small. Don't call a config change a "pivotal moment."
- Never address the reader as "you" in docs unless it's a tutorial with explicit steps. "Users can configure X" not "You can configure X."
- Never use collaborative language. "Let's explore" and "we'll dive into" are chatbot patterns. State the information directly.
- Be specific. "Reduces API calls by 40%" beats "significantly improves performance." Numbers over adjectives.
- Be direct. "This function returns null on failure" beats "This function is designed to gracefully handle failure scenarios by returning a null value."
Verification
After writing any text, check:
- [ ] No banned words/phrases from the list above
- [ ] No Rule of Three patterns (3 adjectives, 3 phrases in a row)
- [ ] No em dashes used where commas would work
- [ ] No promotional tone — would a technical writer approve this?
- [ ] No vague claims without specific evidence
- [ ] Sentence case headings, not Title Case
- [ ] Prose over bullet lists where possible
- [ ] No "in conclusion" or "in summary" wrapping sections
Anti-Patterns
- Writing a landing page that sounds like every other AI-generated landing page
- Blog posts that "delve into" topics and find them "multifaceted"
- README files that describe the project as "a robust, comprehensive solution"
- Emails that start with "I hope this message finds you well"
- Social media posts with emoji before every bullet point
04 同仓库更多 Skills
writer-memory
Agentic memory system for writers - track characters, relationships, scenes, and themes
wiki
LLM Wiki — persistent markdown knowledge base that compounds across sessions (Karpathy model)
website-creation
Build production-quality SaaS websites with opinionated design presets. Use when creating any startup website. The user MUST pick a design style before building. Enforces shadcn/ui, Figma design principles, specific CSS values per style, and anti-AI-writing. Load alongside anti-ai-writing skill. LIGHT MODE ONLY.
visual-verdict
Structured visual QA verdict for screenshot-to-reference comparisons
visual-qa-pipeline
Screenshot every page with Playwright, feed to visual QA agent for design evaluation, report results to Slack. Use after every build phase to verify UI matches the chosen design preset. Integrates with the Playwright skill and the website-creation preset system.
verify
Verify that a change really works before you claim completion