Self Eval

This lets Claude evaluate its own responses before sending them back to you. It's basically a self-check step where the model reviews what it's about to say and can catch mistakes, unclear explanations, or missed requirements. Useful when you want higher quality outputs and don't mind the extra latency and token cost of having Claude do a second pass. The tradeoff is straightforward: better answers but slower and more expensive. Works best for complex tasks where getting it right the first time matters more than speed, like code reviews, technical writing, or detailed analysis work.

skills.sh installs: 582
GitHub stars: 29.5K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add alirezarezvani/claude-skills --skill self-eval --agent claude-code

Installs into .claude/skills of the current project.

Files

Self-Eval: Honest Work Evaluation

ultrathink

Tier: STANDARD Category: Engineering / Quality Dependencies: None (prompt-only, no external tools required)

Description

Self-eval is a Claude Code skill that produces honest, calibrated work evaluations. It replaces the default AI tendency to rate everything 4/5 with a structured two-axis scoring system, mandatory devil's advocate reasoning, and cross-session anti-inflation detection.

The core insight: AI self-assessment converges to "everything is a 4" because a single-axis score conflates task difficulty with execution quality. Self-eval separates these axes, then combines them via a fixed matrix that the model cannot override.