Appendices
Appendices#
The main chapters provide the shortest path through the book and the course project. The appendices preserve deeper methods, implementation details, and research cases that are useful when a particular problem calls for them but are not prerequisites for following the main argument.
Use them selectively:
- Evaluating agentic workflows develops the evaluation vocabulary, experimental discipline, graders, uncertainty, and routing questions introduced briefly in the Benchmarking chapter. It also provides an optional comparison of assistance modes for retained competence.
- Reproducible environments distinguishes repeatable setup, declared environments, hermetic execution, and reproducible builds, and provides optional browser measurement procedures.
- Advanced web boundaries examines state continuity, transactions, replicated state and the CAP trade-off, client activation, constructive invariants, runtime-generated interfaces, interaction styles, executable HTTP contracts, event-view rebuilding, and endpoint/cache policy.
- Agent-system security develops control planes, enforcement, traceability, imported capabilities, cross-run state, external contribution policies, and optional exercises on memory, contract completeness, and human review gates.
- Scalability models gives a deeper treatment of workflow reliability and contention models.
- Durable workflows examines retry-safe external effects and recovery across process failure.
Each appendix states the question it helps answer and links its method back to the relevant chapter or project decision. The case studies provide further optional readings, including Temper's approach to executable behavioral contracts.