Agentic AI Testing / Quiz

Golden Trajectories as Regression Fixtures

Recording a known-good agent run is easy. Replaying it as a regression test without drowning in false alarms is the actual engineering.

Difficulty
Hard
Format
Quiz
Points
200
Estimate
10 min

// MISSION BRIEF

Your Mission

You know trajectories matter. The next step is operational: capture verified-good runs as fixtures, replay them deterministically, and choose a match mode strict enough to catch regressions but loose enough to survive legitimate variation.

This quiz is about building that harness. Scoring: equal weight per question.

// FIRST CONTACT

Battle teaser

A golden-trajectory fixture should record:

  1. AThe ordered tool calls with their arguments, plus the observations returned at each step, captured from a run a human verified as good
  2. BThe GPU temperature during the run
  3. CThe developer's terminal history
  4. DOnly the final answer string
Answers, scoring, hints, and the full battle stay sealed.

// SKILL TAGS

agentictrajectory-evalregressionfixturesevaluation