Self-Critique and Improvement Protocol

# Self-Critique and Improvement Protocol

You are an expert in AI output quality systems. Design a self-critique protocol that instructs an AI to evaluate its own work against specific criteria and produce a genuinely better second version — not just a stylistic variation.

## Output to Improve

**Type of output:** [OUTPUT_TYPE]
**First draft (or instruct AI to generate one first):** [FIRST_DRAFT or GENERATE_INSTRUCTION]
**Intended use:** [USE_CASE]
**Audience:** [AUDIENCE]

## Critique Dimensions

Rate the first draft on each dimension (1–10) and explain the score:

1. **Accuracy:** Are all factual claims correct and verifiable? Score: [SCORE]/10 — [EXPLANATION]
2. **Clarity:** Would [AUDIENCE] understand every sentence without re-reading? Score: [SCORE]/10 — [EXPLANATION]
3. **Completeness:** Does this cover everything [AUDIENCE] needs for [USE_CASE]? What is missing? Score: [SCORE]/10 — [EXPLANATION]
4. **Concision:** Is every sentence doing necessary work? What could be cut? Score: [SCORE]/10 — [EXPLANATION]
5. **[CUSTOM_DIMENSION]:** [CUSTOM_CRITERION] Score: [SCORE]/10 — [EXPLANATION]

## Improvement Mandate

Identify the [TOP_ISSUE_COUNT] most impactful improvements — the changes that will move the needle most, not cosmetic edits.

For each improvement, state:
- **Current version:** [WEAK_EXCERPT]
- **Why it fails:** [REASON]
- **Revised version:** [STRONGER_EXCERPT]

## Final Output

Write the complete improved version incorporating all changes. Then provide a before/after score comparison proving the improvement is real.