APPENDIX H: RESUME PORTFOLIO BUILDING CAPSTONE PROJECTS

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS These two prompts reflect work a harness engineer does when designing evaluations. Replace each bracketed placeholder with your own material before use. [START OF PROMPT] I am writing an evaluation dataset for an agent harness that does this job: [DESCRIBE THE JOB IN TWO OR THREE SENTENCES]. Currently I have these test cases: [LIST YOUR CASE TITLES]. Review the set for coverage. Tell me which kinds of situations are missing, such as empty inputs, very long inputs, conflicting information, unusual formatting, and attempts to break the rules. Propose ten new invented test cases, each with an input description, the expected behavior, and which of my scoring checks should catch a failure. Use only invented data. Output format: Coverage Gaps, then ten cases with the fields Input, Expected Behavior, and Check That Catches Failure. [END OF PROMPT] [START OF PROMPT] Below are the results of my last evaluation run, with pass counts for each check and each model version: [PASTE YOUR RESULTS SUMMARY]. Act as an analyst. Explain in plain language what the numbers do and do not show, identify the two most important things to investigate first, and suggest a fair way to decide whether one version is really better than another when the dataset is small. List any claims I should avoid making in a resume or interview based on these numbers alone. Output format: What The Numbers Show, What They Do Not Show, Investigate First, Fair Comparison Method, Claims To Avoid. [END OF PROMPT] PROJECT 5: THE SUPPORT DESK KITCHEN (A PRODUCTION-HARDENED MULTI-AGENT SYSTEM)