m1nga/write-judge-prompt
Description and use
Agent skill that turns one subjective failure mode and real labeled traces into a binary Pass Fail LLM evaluator prompt.
Compatible agents
- Claude Code
- Hermes
- Codex CLI
Limitations and review notes
GitHub-discovered skill metadata only; install commands are not executed by AtlasRepo.
- Source: GitHub
- License: MIT
- Category: agent-skills
- Record date: 2026-08-26T14:15:33.248Z