← ClaudeAtlas

ritual-voicelisted

Audit written content against a brand's voice guardrails defined in a ritual.config.json file. Use this skill whenever the user asks to review, check, audit, lint, or enforce brand voice on any content surface: landing pages, blog posts, testimonials, case studies, emails, social posts, or marketing copy. Also trigger when the user mentions voice lint, brand guardrails, hype words, AI slop, stale stats, unverified claims, or asks whether copy sounds on-brand. Works across any brand that has a ritual.config.json in the repo. Six priority checks: stale metrics, missing specificity, AI-slop markers, banned hype words, name/attribution mismatches, and generic corporate voice. Part of Ritual: pair with Claude Code routines for scheduled enforcement across repos.
whystrohm/ritual · ★ 0 · Data & Documents · score 67
Install: claude install-skill whystrohm/ritual
# Ritual: Voice Audit written content against a brand's voice guardrails. This skill is brand-agnostic. It reads rules from a `ritual.config.json` file in the repo and applies them to content. Any brand can use it by dropping in their own config. Part of **Ritual**: routines run, Ritual proves. Make every claim earn its place. ## When to trigger Trigger when the user: - Asks to "check," "audit," "review," "lint," or "enforce" brand voice - Mentions a specific piece of copy and asks if it's on-brand - Asks about hype words, AI slop, stale stats, or voice guardrails - Drops a URL, markdown file, or block of copy and asks for a voice pass - Is preparing content to publish and wants a sanity check first Do not trigger for: - Generative writing requests ("write me a blog post"): this is an audit skill, not a writing skill - General proofreading (spelling, grammar): this is voice-specific - Questions about voice strategy or positioning: this enforces voice, doesn't design it ## Three modes: always ask which one Every invocation, ask the user which mode they want: 1. **Flag**: List violations with file + line + rationale. Don't edit anything. 2. **Suggest**: Flag violations AND propose a rewrite in the brand's voice. User decides whether to apply. 3. **Fix**: Apply rewrites directly and show a diff. For use when the user trusts the skill or is running it as part of a routine. If the skill is invoked from a routine with no user present, default to **Suggest** and emit the o