recursive-improvement

A pattern for generating higher-quality AI output by iterating against explicit scoring criteria.

The Pattern

generate → evaluate → diagnose → improve → repeat (until passing)

Most people prompt once and ship whatever comes back. This skill works differently — it generates, evaluates against a scoring checklist, diagnoses weaknesses, rewrites, and re-evaluates. It keeps looping until the output actually hits the bar.

Quick Start

1. Copy SKILL.md into your AI's context (system prompt, project docs, or paste directly)
2. Pick the relevant criteria for your use case (or define your own)
3. Set minimum thresholds (usually 8/10)
4. Let the AI iterate until all criteria pass

Included Criteria

The skill includes ready-to-use scoring criteria for:

• Social content — hooks, engagement, thumb-stop power
• Landing pages — headlines, value props, CTAs
• Email copy — subject lines, skimmability, single focus
• Ad copy — pattern interrupts, emotional triggers, platform fit

Plus a framework for building your own criteria for any repeatable task.

When to Use

✅ Headlines and hooks
✅ CTAs and value props
✅ Landing page copy
✅ Social posts and threads
✅ Ad copy
✅ Important emails

❌ Internal notes
❌ First-pass brainstorming
❌ Technical docs

The Secret Sauce

The real power is in adversarial pressure — after the output passes your criteria, you attack it:

• "Would a skeptical customer believe this?"
• "Would a distracted scroller stop for this?"
• "How would a competitor tear this apart?"

If it survives the attack, ship it. If not, iterate.

---

See SKILL.md for the full implementation.