Terminal-Bench-Science: Evaluating AI agents on research workflows across scientific domains
Latest release v0.1.0 · 26 Aug 2026
5 files for Claude Code, Codex and OpenCode: convert-separate-verifier, review-task, update-rubric, terminal-bench-science AGENTS.md and 1 more — 17,369 tokens loaded in every session.
AGENTS.md A 17,366 tok CLAUDE.md A 3 tok .claude/skills/convert-separate-verifier/SKILL.md A 63 tok .claude/skills/review-task/SKILL.md A 26 tok .claude/skills/update-rubric/SKILL.md A 20 tok These files are harbor-framework/terminal-bench-science's own configuration — they tell Claude Code, Codex and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.