.venv/
__pycache__/
*.egg-info/
.pytest_cache/
dist/
.DS_Store

# benchmark run artifacts. The graded verdicts and the summary ARE the evidence
# for the numbers in BENCHMARK.md, so they ship; everything else here is
# regenerable bulk — raw agent transcripts, per-turn and superseded verdicts,
# the 8MB of answers, and the budget ledger.
benchmark/results/*
!benchmark/results/latest
benchmark/results/latest/*
!benchmark/results/latest/judge
benchmark/results/latest/judge/*.turn*.json
benchmark/results/latest/judge/*.attempt*.json
!benchmark/results/latest/summary.md
!benchmark/results/latest/summary.json
