eval-harness

A framework for defining and running formal evaluations of Claude Code sessions to measure reliability and prevent regressions.

Lift—not yet benchmarked

Trustsecurity · quality · license · upkeep

Scan Saferisk 3/100

LicenseMITserved via MCP

What it does

Formal evaluation framework for Claude Code sessions implementing eval-driven development (EDD) principles. Use when a Claude Code workflow needs a formal eval before it is trusted or changed.

#eval-driven-development#claude-code#passk#regression-testing#agent-evaluation#quality-assurance

Quality judgment

completeness

8.5/10

clarity

9.0/10

trigger quality

7.0/10

examples

8.0/10

structure

9.0/10

overall

8.5/10

Well-organized, clear EDD framework with concrete grader types, metrics, workflow, and a full worked example. Triggers are reasonable but somewhat abstract; some commands like /eval are assumed rather than defined.

allowed-tools

ReadWriteEditBashGrepGlob

Files (3)

agents/openai.yaml · 0.2 KBLICENSE · 1.0 KBSKILL.md · 5.5 KB