Read references/building-production-ready-ai-agents-with-openai-s-agents-sdk-techniques-tips-and-methodology_20251231_043933.md before answering; treat it as the authoritative source of truth for all technical guidance in this skill.
Identify the user's target workflow (single-agent vs multi-agent) and map the request to the relevant reference sections (Agents SDK Foundations and Architecture (foundation_architecture), Designing Agents: Methodology and Patterns (agent_design_methodology), Tooling, Integrations, and Reliability Engineering (tooling_reliability), Evaluation, Testing, and Debugging Agents (evaluation_testing), Deployment, Monitoring, and Governance (deployment_governance)).
Extract the specific SDK primitives, patterns, and methodological steps from the reference that apply; quote or cite the relevant passages/section headings when presenting recommendations.
Specify tool integration and reliability requirements (schemas/constraints, timeouts/limits, validation/sandboxing/permissions, guardrails and handoffs) strictly according to the reference; avoid adding unstated mechanisms.
Define an evaluation and operations plan based on the reference (test strategy, tracing/debugging approach, regression coverage, monitoring/alerts, cost/latency controls, versioning/rollout, governance and human oversight), and ask clarifying questions when the reference-dependent details are insufficient.

build-production-agents-sdk