tov-guidelines
Skill70 starsChanged 11 days agoMITView on GitHub
Extracts a company's tone of voice from its own content and turns it into rules a writer or an AI can follow. Phase 1 analyses 15 to 20 pages for sentence, paragraph, word and structure patterns with frequency counts; after the user reviews it, Phase 2 writes the guidelines: tone attributes with before and after examples, a We are / We are not table, tone by context, vocabulary to use and avoid, and open questions. Triggers: "tone of voice", "brand voice", "voice guidelines", "writing style guide", "extract voice from our website", "TOV". Not for visual identity (use /brand-kit) or messaging (use /messaging).
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
When content has drifted, or the existing guide is more than six months old.
- Very specific: line 25
Don't use it for colours and fonts (/brand-kit), for what to say (/messaging), or when the user only wants a one-line description of the …
- Very specific: line 33
Paths: Write context/brand/MMYY-tov-analysis.md (Phase 1) and context/brand/MMYY-tov-guidelines.md (Phase 2).
- Distinctive: line 86
No adjective-only rules. "Friendly" isn't a guideline.
- Distinctive: line 87
Every example is real text from the site.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
TOV guidelines
Find out how a company actually writes, then write that down as rules specific enough that a new writer, or an AI, sounds like them on the first draft. Two phases with a review in between: first the evidence, then the guidelines.
When to use it
- Building the context folder, before any copy or content work.
- Starting a content program, or onboarding writers or AI tools.
- When content has drifted, or the existing guide is more than six months old.
Don't use it for colours and fonts (/brand-kit), for what to say (/messaging), or when the user only wants a one-line description of the voice.
Inputs
Required: the company website URL. It needs a homepage and at least 5 to 10 pages of real content. If it's thin, say so and flag the sample size in the output.
Optional: founder notes or transcripts, an existing style guide to update, context/icp/ for the reader, context/expert-pov/ for the founder's own voice.
Paths. Write context/brand/MMYY-tov-analysis.md (Phase 1) and context/brand/MMYY-tov-guidelines.md (Phase 2). Agency or consultant? Swap context/ for clients/<client>/context/ throughout.
Tools. Page fetching for the site. Web search with a site filter helps collect founder posts. Exa or Firecrawl are fine if you have them.
Phase 1: analysis
- Pick the pages. Homepage, about or values, 3 to 5 blog posts, 2 to 3 case studies, pricing, FAQ, a landing page. Aim for 15 to 20.
- Sentence patterns: average length, first or third person, questions, imperatives.
- Paragraph patterns: length, how they open, where evidence sits.
- Word patterns: the company's own terms, the customer's terms, words they avoid, how often they use modifiers.
- Structure patterns: sentence or title case headers, CTA wording and placement, how they stack proof.
Score frequency.
- High: 80% of pages or more.
- Medium: 40 to 79%.
- Low: under 40%.
- Conflict: pages contradict each other.
Show the count ("5 of 6 pages").
- Map voice by content type. Person, formality and CTA style for homepage, blog, case study and pricing.
- Collect real examples for each pattern. Use real sentences only, never invented ones.
- Flag inconsistencies and gaps, with the questions you'd ask the founder.
- Review gate. The user confirms the patterns, corrects anything wrong and answers the gap questions. Don't start Phase 2 without this.
Phase 2: guidelines
- Apply the feedback and resolve each inconsistency.
- Write the core guide:
- who reads it
- 3 to 5 tone attributes, each with a before and after rewrite
- vocabulary to use and avoid
- structure templates for the main content types
- We are / We are not table. 5 to 8 rows, each backed by Phase 1 evidence.
- Tone by context. How the voice shifts for these contexts:
- cold outreach
- nurture email
- objection handling
- sales follow-up
- support reply
- case study
- social post
- Tone by reader emotion. How to write to someone frustrated, sceptical, excited or rushed.
- AI writing tells to avoid. At least five patterns that would make this brand sound machine-written, shown with the brand's own fix. Examples: generic praise openers, three short parallel sentences, "it's not X, it's Y" contrasts, filler transitions, closing on "what do you think?".
- Optional sections, only when they apply:
- hard questions answered: security, pricing, a big competitor. Mark these draft until the founder approves.
- product microcopy, for companies with a product UI.
- AI assistant voice, for companies with a chatbot.
- Open questions. At least three, each saying what it unblocks.
- Source every rule to a page, or mark it "TBD, needs founder input".
What good looks like
- At least 15 pages read, with findings in every pattern category.
- Frequencies are counts, never "usually" or "often".
- No adjective-only rules. "Friendly" isn't a guideline; "use 'you' in the first sentence of every page" is.
- Every example is real text from the site.
- A writer could pick the right tone for a support reply without asking anyone.
Mistakes to avoid
- Writing the voice you wish they had instead of the one they have. Note the gap; don't paper over it.
- Skipping the review gate.
- Copying another company's style guide structure and filling it with adjectives.
- Rules so vague any brand could use them.
Chain
Part of the free five-skill chain: /icp-research → /competitor-research → tov-guidelines → /positioning → /messaging. The other six context skills (company context, win-loss, founder POV, brand kit, funnel and goals) are in the paid context plugin ($19.99, link in the README).
Other versions3 projects use a version of this.
Word for word · 1
- matteotitta/repo-structure-in-house★ 70100% same
Edited · 1
- matteotitta/claude-marketing-context★ 097% same
Wrote their own version · 1
- matteotitta/genesys-skills★ 46same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository8 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- context/CLAUDE.mdCLAUDE.md
- outputs/CLAUDE.mdCLAUDE.md
- competitor-researchSkill
- icp-researchSkill
- levelSkill
- messagingSkill
- positioningSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

