{"slug":"minimax-ai-harness-experiment","source_name":"minimax-ai/harness-experiment","name":"Minimax AI/Harness Experiment","description":"Run a harness change as a controlled experiment, with a small end-to-end eval, one change at a time, repeated runs, and looking at flipped tasks, request count, tokens and cache hits, not just the average score. Use before and after changing prompts, tools, skills, subagent layout, compaction or other agent scaffolding, or when asked whether a change helped.","version":1,"lift":{"pass_rate_delta_pts":9.09,"pass_rate_pct":81.8,"total_cases":22,"passed_cases":18,"tokens_delta_pct":72.8,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.8-flash","grading_method":"judged","completed_at":"2026-11-03T00:18:54.347297+00:00"},"skill_score":0.8182,"benchmark_models":[{"model":"gemini-3.8-flash","headline":true,"delta_pts":9.09,"with_pass_pct":81.8,"without_pass_pct":72.7,"tokens_delta_pct":72.8,"turns_delta_pct":0,"total_cases":22,"cases_aggregated":22,"verdict":"mixed","never_hurt":false,"completed_at":"2026-11-03T00:18:54.347297+00:00","run_id":"856630c2-6ef2-45fe-a16a-ec0063f27241","latest_attempt":{"id":"856630c2-6ef2-45fe-a16a-ec0063f27241","runner_model":"gemini-3.8-flash","overall_verdict":"mixed","completed_at":"2026-11-03T00:18:54.347297+00:00","gate":null},"version_number":1,"is_latest_version":true,"gate":null}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"MIT","install_count":0,"total_uses":0,"total_users":0,"manifest_hash":"11f07e56fae246f71a0ef67c455695d6d57a1157841538cb7919b80c6f7a18c6","install_name":"minimax-ai-harness-experiment","version_id":"3e835c68-e3ed-4078-b61c-3897983c81ae","body_sha256":"11f07e56fae246f71a0ef67c455695d6d57a1157841538cb7919b80c6f7a18c6","render_hash":"afb5980743b1006403287a68ab7efb43bb0445d40b2fbb160fed56425a918072","raw_url":"https://app.decimal.ai/s/minimax-ai-harness-experiment/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/minimax-ai-harness-experiment"}