content-quality-checklisted

Editorial QA pass that catches AI tells, template language, hollow hedges, weak closers, and brand-voice violations in any draft before it publishes. Runs a deterministic linter (slop phrases, LLM-overrepresented words, "not just X, it's Y" reframes, em-dash chaining, rhetorical-question openers, announcement cliches, the brand's own Avoid list) and then an editorial judgment pass (throat-clearing openers, closers that restate the open, claims without proof, specificity, voice-pillar conformance). Reports findings by severity with line references and a fix per flag; rewrites only on request. Use whenever the user says "review this draft", "QA this post", "check for AI tells", "does this sound AI-written", "editorial review", "content quality check", "de-slop this", or is about to publish copy. Enforces the standard that brand-voice-guide builds; works standalone with the universal checklist when no voice directory exists.
esmefong/gtm-skills · ★ 0 · AI & Automation · score 72

Install: claude install-skill esmefong/gtm-skills

# Content Quality Check The QA gate a draft passes before it publishes. It exists because of one fact: **readers now pattern-match AI tells in seconds, and every tell that ships spends the brand's credibility.** The check exists to catch what the writer can no longer see in their own draft. It reports before it rewrites. The deliverable is a severity-tagged review with a fix per flag; wholesale rewriting happens only when asked, because the writer owns the draft. ## What this consumes | Input | Why | If missing | |---|---|---| | The draft | The thing under review | Required | | Channel and audience | Calibrates length, register, and which rules bind hardest | Ask in one line | | Voice directory (optional) | `brand-voice-guide` output: the pillars to check conformance against and the Avoid list the linter enforces | Run with the universal checklist only, and say so | The division of labor across the repo holds: `brand-voice-guide` builds the standard, this skill enforces it on drafts, and `outbound-engine` carries its own sequence-specific gate (this skill can still review sequence copy as prose). ## The workflow Four phases. Mechanical first, judgment second, so the model's attention goes where the machine cannot. ### Phase 1: Intake and calibration Read the draft, note channel and audience, and locate the voice directory if one exists (pillars from `voice-core.md`, banned words from the Avoid section of `writing-standard.md`). State the bar being applied: universal