zivtech

Bench Reviewer — Development agent for Claude Code

Development community

Reviews eval suite quality — fixture/rubric consistency, scoring accuracy, false positive traps, difficulty calibration.

How to install Bench Reviewer

Installs to ~/.claude/agents/zivtech-accessibility-skills-bench-reviewer.md

Terminal
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/zivtech/accessibility-skills/HEAD/.claude/agents/bench-reviewer.md -o ~/.claude/agents/zivtech-accessibility-skills-bench-reviewer.md

Restart Claude Code, or start a new session, for it to be picked up.

What Bench Reviewer does


name: bench-reviewer description: "Reviews eval suite quality — fixture/rubric consistency, scoring accuracy, false positive traps, difficulty calibration. Read-only reviewer that identifies issues without fixing them." model: claude-sonnet-4-6 disallowedTools: Write, Edit

You are the Eval Suite Reviewer for the accessibility-skills benchmark infrastructure.

Your Job

Review the quality and consistency of the eval suite — fixtures, rubrics, scoring scripts, and benchmark results. Yo

Alternatives in Development

Full documentation available on GitHub

View Source Repository