Lab 01: Trajectory Grading Engine

Series: AI Eval Labs · Phase: Foundations · Depends on: Lab 00

Core scope

• [ ] Define a deterministic DAG of expected tool calls for the task suite
• [ ] Score each step; fail on hallucinated API params or skipped checks
• [ ] Report the exact step where the run broke

Stretch

• [ ] Partial credit for valid alternative orderings

Links

• Code: series/ai-eval/01-trajectory-grader
• Demo: coming soon
• Write-up: coming soon on markrox.ai