{"slug":"brycewang-stanford-colm-reproducibility","source_name":"brycewang-stanford/colm-reproducibility","name":"Brycewang Stanford/Colm Reproducibility","description":"Use when hardening a COLM paper's reproducibility story — pinning open-weight checkpoints and tokenizers, handling API-model drift and deprecation honestly, versioning evaluation harnesses and prompts, disclosing compute, and writing availability statements that distinguish what is releasable from what is not.","version":1,"lift":{"pass_rate_delta_pts":0,"pass_rate_pct":86.4,"total_cases":22,"passed_cases":19,"tokens_delta_pct":37.3,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-08-31T04:37:16.059004+00:00"},"skill_score":null,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":0,"with_pass_pct":86.4,"without_pass_pct":86.4,"tokens_delta_pct":37.3,"turns_delta_pct":0,"total_cases":22,"cases_aggregated":22,"verdict":"mixed","never_hurt":false,"completed_at":"2026-08-31T04:37:16.059004+00:00","run_id":"39fe98bf-61c7-4c93-87a3-9b02e569ccc9","latest_attempt":{"id":"39fe98bf-61c7-4c93-87a3-9b02e569ccc9","runner_model":"gemini-3.6-flash","overall_verdict":"mixed","completed_at":"2026-08-31T04:37:16.059004+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":"da6a548f1142ea576dfae0d42dc65445f586ffbe172ddd31ba9e3801d5926747","install_name":"brycewang-stanford-colm-reproducibility","version_id":"a5e89cf1-1454-40a3-a469-e93a7355550f","body_sha256":"da6a548f1142ea576dfae0d42dc65445f586ffbe172ddd31ba9e3801d5926747","render_hash":"74298b3214cce63d98c574d23be4c71f05a7f3d8cd94faccf389a769b39883f5","raw_url":"https://app.decimal.ai/s/brycewang-stanford-colm-reproducibility/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/brycewang-stanford-colm-reproducibility"}