Module 08 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. Permission violations should be tracked as…
- A. a rate, like other quality metrics
- B. a count with a threshold of zero
- C. a percentage of runs
- D. an average per week
2. A case set built only from requests people wanted fulfilled…
- A. is representative
- B. never tests refusal, escalation or clarification
- C. is the recommended starting point
- D. over-tests the refusal path
3. Scoring the path as well as the outcome catches…
- A. nothing extra
- B. correct answers reached expensively or via actions nobody asked for
- C. only latency problems
- D. model version changes
4. Removing a case from the set because it fails…
- A. is good hygiene
- B. turns the set into a description of what the agent does rather than a specification
- C. is required when the case is hard
- D. improves representativeness
5. The release decision after a change should cite…
- A. the headline score
- B. the list of cases that used to pass and now fail
- C. the average across runs
- D. the number of cases
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - a count with a threshold of zero
A rate on a busy day rounds one breach to zero, and an alert threshold above zero says some breaches are acceptable.
2. B - never tests refusal, escalation or clarification
Those three behaviours are most of what makes an agent safe to deploy.
3. B - correct answers reached expensively or via actions nobody asked for
Redundant calls and unexpected tools are both computed mechanically from the recorded path.
4. B - turns the set into a description of what the agent does rather than a specification
Once is reasonable if the case was badly specified. As a habit, the score only ever goes up.
5. B - the list of cases that used to pass and now fail
A change that fixes one case and breaks another leaves the total untouched.
Practical Check
Apply this module to your own work: complete the module activity for *Evaluating an Agentic Workflow*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.