content-voice · diff
git:20260705.1efdd2c to git:20260905.0fd23a5
57 added, 121 removed. Audit A to A.
---
name: content-voice
title: "Human Writing Voice"
- description: "AI pattern avoidance, dead vocabulary, voice rules, and rhythm guidance for writing that reads as genuinely human. Apply to any content writing or rewriting task."
+ description: "Professional voice, clear prose, and context-aware editing for posts, replies, articles, and scripts. Preserve author facts and register without formulaic rhetoric or fabricated experience."
license: Apache-2.0
- compatibility: "Octomind content agents."
+ compatibility: "Text editing and source material; network access for current policy or factual checks."
domains: content
rules:
- content(humanize)
- content(voice)
- content(article)
- content(blog)
- match(\brewrite\s+(this|the|my)\s+(article|blog|post|copy|content|draft|piece|writing)\b)
- match(\b(write|writing|draft|drafting)\s+(an|a|the|this)\s+(article|blog|post|copy|piece|essay|newsletter)\b)
- match(\b(make|sounds?)\b.*\b(human|natural|authentic|less\s+ai|less\s+robotic)\b)
- match(\b(ai|gpt|llm)[-\s]?(generated|sounding|tone|copy|writing|text|prose)\b)
- match(\bdead\s+vocabulary\b)
- match(\b(content|copy|article|blog|writing)\s+voice\b)
- match(\b(tone\s+of\s+voice|brand\s+voice|writing\s+voice)\b)
- semantic(write this in a more human voice)
- semantic(make this copy sound natural and authentic)
- semantic(fix the tone of this writing)
- semantic(rewrite this article so it does not sound robotic)
- semantic(remove AI patterns from this draft)
- semantic(polish this content to read like a real person wrote it)
---
## Overview
- AI-generated text has measurable signatures that readers and search engines detect. This skill encodes the complete ruleset for eliminating those signatures — dead vocabulary, uniform rhythm, missing contractions, structural uniformity — and replacing them with patterns that characterise skilled human writing.
-
- Apply this skill whenever writing, rewriting, or editing content intended to read as human.
-
- ## Instructions
-
- ### Dead Vocabulary — Never Use
-
- These words appear 5–10× more in AI text than human text. Replace every instance.
-
- Dead verbs: delve, embark, elevate, unleash, harness, unlock, unveil, foster, navigate, streamline, leverage, underscore, garner, revolutionize
-
- Dead adjectives: profound, innovative, transformative, seamless, meticulous, vibrant, robust, cutting-edge, pivotal, intricate, comprehensive, groundbreaking, crucial, vital, essential, poignant, renowned
-
- Dead nouns: landscape, tapestry, realm, journey, testament, synergy, underpinnings, dynamic (as noun), aspect (overused), complexity (overused)
-
- Dead adverbs: seamlessly, significantly, notably, relentlessly, tirelessly, profoundly
-
- Dead transitions: furthermore, moreover, additionally, consequently, importantly, notably, indeed, essentially, alternatively, ultimately
-
- Dead phrases:
- - "it's important to note"
- - "in today's [anything] world"
- - "in the realm of"
- - "designed to enhance"
- - "diverse array"
- - "rich tapestry"
- - "ever-evolving"
- - "game changer"
- - "it's worth noting"
- - "delving into the intricacies"
- - "navigating the complexities"
- - "a testament to"
- - "not just X, but also Y" — rewrite this construction every time
- - "what nobody tells you" / "what no one else is talking about" — itself a cliché now, don't open with it
-
- Second-wave vocabulary (2025–2026) — the first wave (delve, tapestry, realm) faded once everyone knew to grep for it; models moved on to these. Refresh this list as new tells emerge — vocabulary rotates, structure doesn't:
- - "quiet" / "quietly" as a faux-profound modifier — "a quiet shift", "quietly reshaping" (literal quiet — a quiet room, a quiet voice — is fine)
- - "drift" / "drifting" as an abstract metaphor — "cultural drift", "we're drifting toward X"
- - "signal" as a vague noun for "sign" or "indicator" — "a signal that things are changing" (a technical/data signal is fine)
- - "cascade" as a metaphor for chain effects — "a cascade of consequences" (a literal cascade — water, CSS — is fine)
-
- Formal overreach — default to the word you'd say out loud. If you'd hesitate to use it in conversation with a friend, swap it for the simpler one. Plain words win almost everywhere; reach for the formal version only when it carries meaning the simple word can't:
- - "utilize" → "use"
- - "facilitate" → "help"
- - "demonstrate" → "show"
- - "commence" → "start"
- - "regarding" → "about"
- - "leverage" → "use"
- - "implement" → "do" / "build" / "run"
-
- Trailing participles — never end sentences with these:
- - "...highlighting the importance of"
- - "...ensuring that"
- - "...contributing to"
- - "...underscoring the need for"
- - "...demonstrating that"
-
- Rewrite as a separate sentence instead.
-
- ### Voice Rules
-
- Contractions — always
- "don't" not "do not", "it's" not "it is", "won't" not "will not", "can't" not "cannot". Uncontracted forms read as AI or legal text. More than 2 uncontracted forms per 500 words is a detectable signal.
-
- Sentence rhythm — dramatic variation
- Never write 3 consecutive sentences of similar length. Follow a 30-word analytical sentence with a 5-word one. Use fragments for emphasis. One-sentence paragraphs between dense ones. This is burstiness — humans write it naturally, AI doesn't.
-
- Good rhythm:
- > The algorithm processes every request through three validation layers before committing to state. That's by design. A single bad write corrupts the entire chain, and recovery is expensive.
-
- Bad rhythm (AI):
- > The algorithm processes every request through three validation layers. This ensures data integrity throughout the system. The design prevents corruption and reduces recovery costs.
-
- Don't rotate synonyms
- If "users" is the right word, keep saying "users." Artificial variation ("users/customers/clients/patrons") across paragraphs reads as AI. Pick the right word and stay with it.
-
- Don't reach for three
- Groups of three adjectives or phrases ("bold, innovative, and transformative") are a rhythm crutch, not a real pattern of thought. One precise word beats three approximate ones. Flag any triad and cut it to the single word that's actually true.
-
- Take a position
- Hedging or both-sides framing — "on one hand... on the other", "it depends", "there are pros and cons to both" — reads as AI avoiding commitment. If a real position exists underneath the hedge, state it. Uncertainty is fine when it's genuine ("the data is mixed here"); false balance when there's actually a right answer is not.
-
- Personal stake
- Occasional first-person: "I've seen this pattern in...", "What surprised me was...". Acknowledge genuine uncertainty: "The data is mixed here." Show emotional range: enthusiasm, skepticism, curiosity. These are human signals.
+ Write clear, useful content in the actual author's voice. Apply this to drafting and revision across posts, replies, articles, and scripts; platform skills provide local format and participation rules. These are editorial practices, not a test of whether a human or AI wrote something.
- Natural grammar breaks
- Start sentences with "And" or "But" when it flows. Use parenthetical commentary. Occasional fragments for emphasis. These are human patterns — don't avoid them.
+ ## Mental model
- Active voice
- "Studies show X" not "X has been shown by studies." Flag passive constructions and rewrite to active unless passive is intentional for emphasis.
+ Professional voice comes from a definite point, credible detail, and respect for the reader. Natural writing can be formal, conversational, technical, or concise. It does not require slang, mistakes, dramatic rhythm, or a story about the speaker.
- Em-dashes — exception, used sparingly
- A single em-dash is natural human punctuation — people use them when thinking-while-writing. The AI signal isn't the em-dash itself, it's density. Three em-dashes in one paragraph reads as GPT default. Keep total use under ~1 em-dash per 100 words and mix punctuation: full stops, commas, parentheses, occasional em-dash. Don't strip them all out either — total absence is its own tell now (over-corrected humanization).
+ ## Establish the brief
- Format follows the venue
- The shape of the text adapts to where it lives. The same idea looks different in a reply versus a long article, and getting that wrong is one of the loudest AI tells in 2026.
+ - Identify the speaker, intended reader, task, venue, and source material. Use approved samples to understand register, terminology, humor, and level of detail; don't copy another person's identity or invent a persona.
+ - With no voice samples, use clear, restrained prose and the supplied facts. Missing samples do not block an ordinary draft. Missing evidence blocks the unsupported claim.
+ - Preserve numbers, names, quotations, dates, technical terms, conditions, and uncertainty. A polished sentence must not make the claim stronger.
+ - First person is appropriate for experience or decisions the speaker supplied. Researching a product does not authorize “I tested it.” Don't invent customers, struggle, excitement, vulnerability, or quotations to create personality.
- - Replies and comments — prose. 1–4 sentences for most. No headers, no numbered lists, no bullet stacks, no bold mid-sentence. People don't write replies as slide decks; AI does by default. Match the surrounding chain — if everyone is writing 1-line lowercase replies, a 6-line structured response stands out as a bot.
- - Posts and threads — each post is a standalone piece of prose. No "1/12" markers, no "🧵", no thread-promo intros. Write each one like a single considered post.
- - Long articles, blogs, docs — structure is fine when it serves reading. Headers break sections that earn it; bullets handle truly enumerable lists (3+ genuinely parallel items); bold marks the rare phrase that needs marking. The test: strip the formatting — does the reader lose anything? If not, drop it.
+ ## Write for the reader
- Lists, headers, and bold are cheap — they're what AI reaches for whenever it isn't sure what to write. Use them only when they earn their place.
+ - Lead with the useful fact, observation, answer, or decision. A hook should tell readers why the content matters and be paid off by the content; don't manufacture suspense around an ordinary fact.
+ - Give enough context to understand the point without the planning brief. Explain unfamiliar terms once; preserve precise domain vocabulary when the audience needs it.
+ - Prefer concrete verbs and identifiable actors. Replace “a transformative solution” with what it actually does, if known. If specifics are unavailable, narrow the claim instead of supplying plausible details.
+ - Remove throat-clearing, repeated claims, unsupported superlatives, and abstract modifiers. Words such as “innovative,” “seamless,” or “leverage” invite a meaning check, not a mechanical ban. Literal or technical uses may be correct.
+ - Keep terminology consistent. Don't rotate “users,” “customers,” and “buyers” for variety when they name different people.
+ - State an evidenced position clearly; keep real qualifications. “May,” “in this sample,” and “under these conditions” can be essential to accuracy, not weakness.
- ### Transitions
+ ## Rhythm and structure
- Don't use a transition when the logic is clear — just start the next idea. When you need one:
+ - Read for comprehension and natural emphasis. Split an overloaded sentence or join choppy ones when it improves the logic. Don't enforce sentence-length alternation, contraction percentages, fragments, or punctuation quotas.
+ - Contractions, fragments, lowercase, and slang should fit the author's register and venue. Correct accidental errors; never add typos, fake edit notes, broken grammar, or invented dialect as camouflage.
+ - Use active voice when it clarifies responsibility; passive voice is useful when the actor is unknown or the process/result deserves focus.
+ - Treat repeated contrast slogans, padded adjective lists, staccato paragraph stacks, and compulsory morals as review prompts. Keep a real comparison, a meaningful list of three, or a useful summary.
+ - Use questions when their answers matter, not as automatic openers or “Agree?” endings. End after the point, a useful next action, or an unresolved question grounded in the subject; no mandatory twist or inspirational closer.
+ - Use headings, lists, and tables when they help readers navigate or compare. A technical reply may need steps or code; a brief conversational reply usually needs prose. Length and formatting follow the task, not an authorship stereotype.
- ✅ "But." / "The problem is —" / "Here's what that looks like." / "Which raises the question:" / "And this matters because —" / "So what does this mean?" / "The catch?" / "Plus,"
+ ## Platform adaptation and trust
- ❌ Never: "Furthermore," "Moreover," "Additionally," "Consequently," "It's worth noting"
+ - Adapt the same factual core to each venue's reader intent and medium. A narrated demo, a support reply, and a professional announcement need different context and pacing; changing a few synonyms is not adaptation.
+ - Read the parent post and relevant replies before answering. Add a specific answer or respectful correction; skip empty praise, imitation of the speaker, and unrelated promotion.
+ - Give a legitimate CTA only when useful. Asking a real question or inviting a relevant action is different from manufactured engagement, false urgency, or a promised reward that cannot be delivered.
+ - Keep commercial affiliation and required provenance disclosures visible. For US-facing endorsements, FTC guidance requires truthful experience and clear material-connection disclosures; platform controls alone may not communicate the relationship sufficiently. Check applicable local and platform rules.
+ - Keep alt text descriptive and faithful to visible information; retain material chart labels or provide equivalent explanation. Review captions for names, negation, numbers, and timing. Accessibility fields are not keyword dumps or extra sales copy.
+ - For another language, match local register and meaning rather than translating English catchphrases literally. Use contextual evidence or qualified review for uncertain idiom; low search frequency is not proof that a phrase is wrong.
- ### Conclusions
+ ## Examples
- Never open a conclusion with:
- - "In conclusion,"
- - "Overall,"
- - "In summary,"
- - "To summarize,"
- - "To wrap up,"
+ Illustrative source facts: a product exports CSV; importing CSV is not supported.
- End by circling back to the opening, posing a question, making a prediction, or just stopping. A conclusion that recaps all main points is an AI pattern.
+ Before:
+ > This revolutionary update unlocks seamless data freedom.
- ## Examples
+ After:
+ > You can export your data as CSV. CSV import isn't supported.
- ### Dead vocabulary replacement
+ The rewrite adds only the stated capability and limitation. It does not invent a user result.
- ❌ `This innovative solution leverages cutting-edge technology to seamlessly facilitate collaboration.`
+ Illustrative source: the author's team tested one repository; runtime was not measured.
- ✅ `This tool uses a shared workspace so teams can work on the same document without stepping on each other.`
+ Avoid:
+ > We made every workflow dramatically faster.
- ### Rhythm fix
+ Use:
+ > We tested this on one repository. We haven't measured runtime yet.
- ❌ `The system processes requests in parallel. This improves throughput significantly. Users experience faster response times as a result.`
+ ## Checklist
- ✅ `The system processes requests in parallel, which improves throughput. Users notice it immediately. Pages that used to take 800ms now load in under 200.`
+ - [ ] The opening delivers a specific point for the intended reader.
+ - [ ] Every experience, result, and quotation belongs to the stated speaker and source.
+ - [ ] Edits preserve scope, uncertainty, and technical meaning.
+ - [ ] Rhythm, formatting, and vocabulary serve comprehension without artificial quotas.
+ - [ ] Platform participation, disclosure, accessibility, and publication conditions are satisfied.
+ - [ ] The final copy is separate from internal source notes and unresolved questions.
- ### Trailing participle fix
+ ## References
- ❌ `The update ships next week, highlighting the importance of backward compatibility.`
+ Reviewed 2026-09-05. Apply source principles within their stated scope; they do not establish social-platform ranking factors.
- ✅ `The update ships next week. Backward compatibility matters here — anything that breaks existing integrations will be reverted.`
+ - [GOV.UK clear-language guidance](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/writing-guidelines/clear-language/) — a public-service style guide supporting audience-focused clarity, not a universal brand voice.
+ - [FTC endorsement disclosures](https://www.ftc.gov/business-guidance/resources/disclosures-101-social-media-influencers) — US-facing endorsements and truthful experience.
+ - [W3C image alternatives](https://www.w3.org/WAI/tutorials/images/decision-tree/) — communicate an image's purpose and relevant information.
+ - [Google's AI-content guidance](https://developers.google.com/search/docs/fundamentals/using-gen-ai-content) — accuracy and reader value, not guaranteed traffic from “human” style.
+ - [Fact grounding](../content-grounding/SKILL.md) and [structured revision](../content-humanize/SKILL.md).