Iterating Agensi Skills (v1.2)

Optimize and Refine Your AI Agent Skills

The first version of an AI skill is rarely the best. To build truly reliable agents, you need a systematic way to observe performance and fix weaknesses. Iterating Agensi Skills provides a lightweight, developer-centric framework for diagnosing and improving existing SKILL.md files based on real-world usage and feedback.

What it does

This skill guides your AI agent through a multi-phase review process to identify why a skill is failing or underperforming. It moves beyond simple "prompt engineering" by evaluating the entire skill structure, including operating procedures, guardrails, and output templates.

• Structured Diagnosis: Evaluates skills against six core areas: Clarity, Boundaries, Procedures, Output, Safety, and Metadata.
• Evidence-Based Improvement: Prioritizes fixes based on actual session logs and observed errors rather than guesswork.
• Actionable Refinement: Converts vague instructions into strict, multi-phase procedures with hard stop conditions.

Why use this skill?

Manual prompt tweaking often leads to "regression bugs" in agent behavior. This skill provides a repeatable methodology to raise the quality of your entire skill library. It ensures your agents have clear expected outputs and robust guardrails, making them significantly more dependable in production environments.

Output Format

Outputs include a structured Skill Review Summary featuring a prioritized diagnosis table, a list of high-impact changes, and specific "Phase-by-Phase" improvement plans to harden your agent's logic.