Eval log
Skill rulebased-io/harness/rulebased-harness/skills/eval-log
Evaluates a Claude Code conversation log against harness compliance — measures human intervention, agent autonomy, tool usage, and session patternsFrom its SKILL.md
npx -y skills add rulebased-io/harness --skill eval-logAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.0 KB, 196 tokens by cl100k_base, as published. Nobody here has run it
Evaluates a Claude Code conversation transcript against harness engineering compliance.
Analyzes the JSONL session log at ~/.claude/projects/<project>/<session>.jsonl.
Evaluation Criteria
- Human Turn Count: fewer human turns = more autonomous. Ratio of human to total turns.
- Autonomy Ratio: percentage of agent-driven turns. Higher is better.
- Build/Test Execution: Bash tool invocations as proxy for builds and tests.
- Tool Diversity: unique tools used (Read, Write, Edit, Bash, Grep, etc.)
- Session Duration: not too short (incomplete) or excessively long.
How to Evaluate
- Find the session log at
~/.claude/projects/<project-path>/<session-id>.jsonl - Parse JSONL, count human vs assistant turns
- Count unique tools and Bash invocations
- Calculate scores per criterion and overall grade
$ARGUMENTS
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.