grandamenium

Skill Optimizer — Development skill for Claude Code

Development community

A Claude Code skill that audits how well an agent followed a skill — reads JSONL transcripts, scores 5 dimensions, writes improvement diffs.

How to install Skill Optimizer

This entry records only its repository, not the path inside it, so there is no exact command to give. Open grandamenium/skill-optimizer and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Skill Optimizer does

A Claude Code skill that audits how well an agent followed a skill — reads JSONL transcripts, scores 5 dimensions, writes improvement diffs.

Alternatives in Development

Battle — Head-to-head graded battle card between two sites — eight dimensions, bar-by-bar, verdict line 3.6k ★
Custom Dimensions — Custom Dimensions and Event Parameters 242 ★
Tech Debt — Multi-agent technical-debt assessment — fans out region-scoped + specialist hunters, verifies findings, scores 107 ★

README

skill-optimizer

A Claude Code skill that audits how well an agent followed a skill by analyzing the session JSONL transcript.

What It Does

After any skill-driven agent run, point this at the transcript and the skill file. It:

Parses the JSONL transcript to extract what the agent actually did
Reads the SKILL.md to understand what the agent was supposed to do
Scores 5 dimensions (1–10 each): trigger accuracy, step coverage, reference accuracy, execution quality, output match
Writes a findings report with the root cause of any gaps
Generates an exact diff patch to improve the skill for next time
Tracks scores over time in a history file so you can see if the skill is improving

Installation

git clone https://github.com/grandamenium/skill-optimizer ~/.claude/skills/skill-optimizer

Usage

In Claude Code:

Run skill-optimizer on ~/.claude/projects//.jsonl against ~/.claude/skills//SKILL.md

Output

skill-improvement/
  /
    YYYY-MM-DD-HH-analysis.md    ← scores + findings
    YYYY-MM-DD-HH-diff.patch     ← exact suggested changes to SKILL.md
    history.json                 ← trend data across all runs