APPENDIX H: RESUME PORTFOLIO BUILDING CAPSTONE PROJECTS

HOW TO PRESENT IT IN A PORTFOLIO README outline Write a readme file, which is the front page of your project, with these headings in this order. Title and one-sentence purpose. What the harness does, in plain language, with a small diagram of the loop drawn as text. How to run it in under five commands, using the mock model. What the permission gate allows and refuses. How to run the tests and what passing looks like. Limits and honest notes, including the fact that the model is a mock. What I would improve next. What to screenshot or record Take a screenshot of the terminal showing ALL TESTS PASSED. Take a screenshot of run_log.jsonl showing a blocked event beside the sneaky task. Record a short screen video, under two minutes, of you running python harness.py, opening output/summary.txt, and then running the sneaky task. Make sure no screenshot shows your secret keys, your home folder name if you prefer privacy, or any real personal data. Metrics to report honestly Report the number of tests and how many pass, for example three of three. Report the number of steps a normal run takes, which you can count in the log. Report the number of requests blocked in the sneaky run. State clearly that these numbers come from a mock model and invented notes, and do not present them as measures of real-world quality. How to word a resume bullet truthfully Good example: "Built a Python agent harness with a tool registry, a folder-level permission gate, a step limit, and structured logging; wrote automated tests showing out-of-bounds file access is blocked." Another good example: "Implemented a think-act-observe agent loop with a swappable model adapter, tested end to end with a mock model." Avoid wording like "built an AI that reads and understands documents," because the mock only follows a script, and avoid saying "production-ready," because it is not. Interview talking points Explain the think, act, and observe loop using the librarian analogy. Explain why the gate resolves a path to its final form before checking it, and what the climbing-dots trick is. Explain why a step limit exists and what you would do if a real task needed more steps. Explain what a mock model is, why you used one, and how you would swap in a real one. Describe one bug you hit while building and how you found it from the log.