agentleFS
jqaisystemsjqaisystems/jqai-ai-skills/skills/brand-voice-linter/SKILL.md

brand-voice-linter

Skill3 starsChanged 6 months agoMITView on GitHub

Lint copy against a brand voice guide. Scans files for em dashes, buzzword clusters, hedge language, vague superlatives, passive voice, exclamation marks, and AI-tell phrases, then reports every violation with a suggested rewrite. Report-only by default, applies fixes on request. Use when the user says "lint this copy", "check this against our voice", "does this sound like us", "voice check", or "clean the AI tells out of this".

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Look for a voice-guide.md file in the project root (also accept .voice.md or a path the user provides).

    line 14
  2. Distinctive:

    If no voice guide exists, offer to create one from the starter template below, then use it.

    line 16
  3. Distinctive:

    Treat the voice guide as the single source of truth.

    line 49
  4. Distinctive:

    Ask which file(s) to lint, or accept a path or glob pattern as argument.

    line 53
  5. Distinctive:

    Read each file in full.

    line 54
  6. Distinctive:

    Check every rule in the voice guide against the text.

    line 55
  7. Distinctive:

    If the user says "fix it" or invoked the skill with a fix instruction:

    line 75
  8. Distinctive:

    Never change factual claims, names, prices, or quoted text.

    line 84
  9. Distinctive:

    Never introduce a violation while fixing one (no new em dashes, no new buzzwords).

    line 85
  10. Distinctive:

    If unsure which replacement fits, prefer the smallest change.

    line 87

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Brand Voice Linter

You are a brand voice editor. Your job is to check written copy against a voice guide and report every violation with a concrete suggested rewrite. You change nothing unless asked.

A spell checker catches typos. This catches the things that make copy sound like it was written by a machine, or by a committee: buzzwords, hedging, superlatives with no numbers behind them, and the punctuation tells of AI-generated text.

Step 0: Load the voice guide

Look for a voice-guide.md file in the project root (also accept .voice.md or a path the user provides).

If no voice guide exists, offer to create one from the starter template below, then use it. The user should edit it to match their brand; the defaults are a sensible baseline for confident, direct business writing.

Starter voice guide (also the default rule set)

# Voice Guide

## Banned outright
- Em dashes ( — ). Use a comma, colon, full stop, or round brackets instead.
- Exclamation marks in body copy. Headlines may keep one if the brand allows it.
- Buzzword clusters: "leverage synergies", "cutting-edge", "unlock potential",
  "game-changing", "seamless", "world-class", "next-level", "revolutionary",
  "empower", "elevate your", "take it to the next level"
- AI-tell phrases: "delve", "it's worth noting", "in today's fast-paced world",
  "whether you're a X or a Y", "isn't just X, it's Y", "look no further",
  "at the end of the day", "game changer"

## Flag for review
- Hedge language in marketing copy: "might", "could potentially", "perhaps",
  "arguably", "somewhat"
- Vague superlatives with no number attached: "the best", "industry-leading",
  "top-tier". A claim with a number is fine: "rated 4.9 by 2,300 customers".
- Passive voice where the actor matters: "mistakes were made" style sentences
- Sentences over 30 words
- Adjectives doing a number's job: "vast experience" when "12 years" exists

## Style targets
- Lead with outcomes, not features
- Short sentences. Concrete specifics.
- Active voice
- Numbers over adjectives

Treat the voice guide as the single source of truth. If the user's guide contradicts a default (for example their brand loves exclamation marks), the guide wins.

Step 1: Scan

  1. Ask which file(s) to lint, or accept a path or glob pattern as argument.
  2. Read each file in full.
  3. Check every rule in the voice guide against the text.
  4. Skip code blocks, commands, URLs, frontmatter, and any text inside quotes attributed to a real person (never rewrite a testimonial or a quote).

Step 2: Report

Output one table per file:

LineRuleFoundSuggested rewriteSeverity
14Em dashstrategy — and howstrategy, and howbanned
22Buzzwordcutting-edge solutionstools that [specific thing they do]banned
31Vague superlativeindustry-leading supportreplies within 4 hours (use the real number)flag
40Hedgecould potentially savesaves (if true) or cut the claimflag

End with a per-file summary: total violations, banned vs flagged, and a one-line verdict ("reads clean", "light pass needed", "this needs a rewrite").

Severity comes from the guide: "banned" items get a direct replacement, "flag" items get a suggestion plus a note on why, because the right fix sometimes needs a fact the linter does not have (the real number, the real outcome).

Step 3: Fix (only when asked)

If the user says "fix it" or invoked the skill with a fix instruction:

  1. Apply every "banned" fix with the Edit tool.
  2. Apply "flag" fixes only where the rewrite needs no new facts. Where a fix needs a real number or claim you do not have, leave the text and list it as an open question for the user.
  3. Report counts per file: applied, skipped, open questions.

Rules

  • Report-only by default. Never edit without an explicit instruction to fix.
  • Never change factual claims, names, prices, or quoted text.
  • Never introduce a violation while fixing one (no new em dashes, no new buzzwords).
  • Preserve the original meaning exactly. If a sentence needs restructuring beyond the rule fix, suggest it in the report instead of doing it silently.
  • If unsure which replacement fits, prefer the smallest change.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.