ask-text-writing
Text Writing: Writes emails, posts, articles, bios, READMEs, and docs that sound human and avoid AI writing patterns. Use when drafting or rewriting prose. Common triggers: anti-slop, make this sound human, sound human, not AI, does not read like AI, write a tweet, draft email, write an email, cover letter, linkedin post, newsletter, blog post, copywriting, schrijf als mens, niet AI, menselijk laten klinken, text-writing, text writing, schrijf de tekst, schrijf de teksten, herschrijf de tekst, teksten herschrijven.
- License 许可
- MIT
- 最近更新
- 2026-11-09
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- ask-text-writing
- description
- Text Writing: Writes emails, posts, articles, bios, READMEs, and docs that sound human and avoid AI writing patterns. Use when drafting or rewriting prose. Common triggers: anti-slop, make this sound human, sound human, not AI, does not read like AI, write a tweet, draft email, write an email, cover letter, linkedin post, newsletter, blog post, copywriting, schrijf als mens, niet AI, menselijk laten klinken, text-writing, text writing, schrijf de tekst, schrijf de teksten, herschrijf de tekst, teksten herschrijven.
- disable-model-invocation
- true
02 资源文件 1
- 参考文档references/banned-words.md2.7 KB
03 SKILL.md 正文
Human-First Writing
Produce text that does not read as AI-generated. Every piece of text — tweets, emails, articles, reports, messages — follows these constraints. Apply them silently; never mention the rules or the guidelines.
Banned vocabulary
Load and respect the list in references/banned-words.md. Never use any banned word, phrase, or opener. Replace it with a concrete specific alternative or restructure the sentence.
Structure
- No rule of three. Use two, four, one, five. Never three by default.
- Vary sentence length. No three consecutive same-length sentences. Mix 4-word sentences with 30-word ones — the most measurable AI signal.
- No parataxis. Do not chain short declarative sentences. Connect related thoughts with clauses, conjunctions, semicolons, or commas.
- No hedging seesaw. Pick a side; state it plainly. One sentence max for counterpoints.
- No corporate pep talk. Write like someone with experience, including the frustrating parts.
- Vary paragraph shape. Do not always open with topic sentence, then explanation, then example, then transition. Some paragraphs are one sentence; some just stop.
- Active voice. No "is being done," "was found to be." Write direct.
- No "As [role], I..." Say the thing without announcing credentials.
Punctuation
- One em dash max per 500 words — the most cited AI tell. Use commas, semicolons, colons, parentheses.
- One exclamation mark max per 1000 words. Enthusiasm comes from word choice.
- Ellipses only when genuinely trailing off, one per piece.
- Use semicolons and colons; good writers use them naturally.
Be specific
- Concrete beats general: "Three clicks from wallet connect to your first risk score" beats "a seamless user experience."
- Real numbers beat claims: "34 users in week one, 12 returned" beats "significant growth."
- Name real things: "Solana, specifically" beats "various blockchain networks."
- Include friction and mess: "The RPC kept timing out at 3am" beats "a rewarding journey."
- Use contractions: "don't", "can't", "it's".
- Ground in time, place, context: "last Tuesday," "at 2am."
- Fragment or run on when the thought is unfinished. Ugly is human.
- Reach past the first word that comes to mind.
Accuracy
- Never invent data, studies, statistics, or quotes. Say "roughly" or acknowledge uncertainty.
- Take clear positions when evidence is solid; qualify only genuine uncertainty.
- Use real verifiable names, companies, dates: "OakNorth" beats "a major bank."
Formatting
- No markdown headers, bold, or asterisks in emails, DMs, SMS, social. Plain text only.
- No emoji as bullet points, no emoji per line, no "🧵" openers, no hashtag stacks.
- No "Thread:" or "Let's dive in" openers.
Voice calibration
For a specific person, match their voice: swearing, slang, sentence length, humour, what they would never say. Default: direct, slightly informal, contractions, occasionally "And"/"But" starts, does not over-explain.
Self-check before output
- Banned words or openers? Replace.
- Three same-length sentences in a row? Vary.
- Three short declaratives in a row? Merge or connect.
- Grouped in threes? Break it.
- Hedging? Pick a side.
- More than one em dash? Cut.
- Passive construction? Make active.
- Every paragraph ends with a transition? Cut some.
- Fabricated specifics? Remove or mark hypothetical.
- Could any AI have written this for anyone? Add something specific.
Use with
designfor web copy inside a design, thendesign-reviewbefore showing itdesign-reviewto review existing copy for AI-default patternswrite-skillif writing quality reveals a reusable gap worth capturingsession-reviewto file follow-up improvements
Avoid
- Surfacing the rules to the reader — apply silently
- Rule-of-three lists, uniform rhythm, passive voice, invented specifics
- Treating this as a style guide for design — that is
design-review's job
Adapted from jalaalrd/anti-ai-slop-writing (MIT).
04 同仓库更多 Skills
ask-write-skill
Write Skill: Creates and revises skills and guardrails from reusable workflow gaps. Use when making a skill or when agent behavior exposes a routing gap.
ask-verification
Verification: Proves work is fixed or ready and cleans up changed repos. Use before claiming something works or handing off.
ask-text-writing
Text Writing: Writes emails, posts, articles, bios, READMEs, and docs that sound human and avoid AI writing patterns. Use when drafting or rewriting prose.
ask-summary
Summary: Explains what changed and what is proven, walks through code, or re-pitches unclear output. Use when work is done or the user asks what changed.
ask-session-review
Session Review: Reflects on skill usage and gaps and files GitHub issues from bugs, findings, or follow-ups. Use after completing work.
ask-research
Research: Does bounded fact-finding across the repository, docs, code, or external sources. Use when a question needs evidence before an answer or decision.