whieet/harness-kit

用 harness engineering 让 Claude Code 长任务更可靠:计划门、验证循环、独立评估,减少跑偏和漏测。

5Stars on the repository
23Mods indexed here, across every type
1mo agoLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown

evaluator

01

whieet/harness-kit

Agent

Skeptical, fresh-context evaluator for Generator/Evaluator separation. Scores the current change against the project's rubric (.harness/rubric.md) on each dimension 1-5, running the config.verificationRecipe checks. Any dimension <3 = FAIL. Use before declaring a coding task done. It cannot edit files, so it never…

5 1mo ago A 77 tokens original MIT