APPENDIX H: RESUME PORTFOLIO BUILDING CAPSTONE PROJECTS

HOW THE FIVE PROJECTS CLIMB Project 1 is a minimal tool-using agent harness with a permission gate and a log. It relates to Chapter 65 and to the first stage of the Practice Lab, because both build the think, act, and observe loop, but Project 1 uses a notes folder instead of resumes. Project 2 is a context-managed research agent with compaction and a notes file. It relates to Chapter 66 and to the second stage of the Practice Lab, but it answers questions from a set of invented garden documents instead of reviewing resumes. Project 3 is a safe tool agent connected to a small server that speaks the style of the Model Context Protocol (MCP), with defenses against prompt injection and an approval step for risky actions. It relates to Chapter 67 and to the third stage of the Practice Lab, but it uses an invented library catalog as its outside data. Project 4 is an evaluation and observability suite with traces, a dataset, scoring, a quality gate, and a dashboard, applied to a harness. It relates to Chapter 68 and to the fourth stage of the Practice Lab, but it measures a different harness and compares several versions of it side by side. Project 5 is a production-hardened multi-agent system with a supervisor, budgets, retries, a kill switch, an incident runbook, and a deployment plan. It relates to Chapter 69 and to the fifth stage of the Practice Lab, but it triages invented customer support tickets instead of resumes. Each project is a little harder than the one before, and each one reuses some of the ideas or even some of the files of the project before it, so your skills pile up like bricks in a wall.