Guardrails

Overview

This skill scans drafts against patterns that recur in Every pre-publication review. The first four categories apply to any Every draft. Categories 5 and 6 are Working Overtime-specific; do not apply them to other columns or ghostwritten work.

This skill produces a findings report. It does not rewrite the whole draft. Katie wants to see flags with diagnoses and suggested fixes, then make the editorial calls herself.

When this skill auto-triggers

• Reviewing any Every draft for pre-publication
• Polishing a draft when AI tells have been flagged
• Invoked explicitly via /guardrails
• Any Every draft review request that mentions "scan," "check," "review," "tighten," or "flag"

For generic AI-tell detection, use ai-check. For Every house mechanics, use every-style. For residual process narration, use tracks. All four can run in sequence; this skill is the substantive editorial guardrail layer.

Detection categories