CedarCopilot/cedar-mail

Branch: staging

test(eval): three instruction-eval harnesses and their negative results

closed#2714CedarCopilot

CedarCopilot wants to merge 2 commits into staging from chore/f8-eval-harnesses

Libra does not show diffs here. View the full diff on GitHub.

addedapps/server/.claude/skills/playbook-instruction-eval/aop-classification-eval.ts+529-0
addedapps/server/.claude/skills/playbook-instruction-eval/deal-close-task-cancel-eval.ts+696-0
addedapps/server/.claude/skills/playbook-instruction-eval/domain-return-shape-eval.ts+1151-0
addedapps/server/.claude/skills/playbook-instruction-eval/fixtures/aop-classification-cases.ts+480-0
addedapps/server/.claude/skills/playbook-instruction-eval/fixtures/aop-selection-baseline.md+212-0
addedapps/server/.claude/skills/playbook-instruction-eval/fixtures/deal-close-next-steps-baseline.md+37-0
addedapps/server/.claude/skills/playbook-instruction-eval/fixtures/deal-close-task-cancel-cases.ts+563-0
modifieddocs/design/f8-agent-log.md+61-0