This is the companion piece to Scaling Agentic AI — if Volume I diagnosed the disease, this is the prescription. The nine engineering disciplines that separate a demo agent from one that survives in production, written as a build manual rather than a survey: each section identifies the discipline, explains the failure mode it addresses, and gives the practical pattern that works. Together with Volume I, the two-part series is meant to be the document I wish someone had handed me before I shipped my first agent system.
What it covers
Nine disciplines, about twenty minutes of reading.
§ 1 — Context engineering. What goes in the prompt, in what order, with what compression. The single most important production discipline.
§ 2 — Tool surface design. Why the function signature you expose matters more than the model you call. Typed tools, parameter constraints, return-value discipline.
§ 3 — Orchestration patterns. Sequential pipeline, supervisor-worker, parallel fan-out, hierarchical. When each one is the right shape.
§ 4 — Observability. Tracing, logging, replay. Why you cannot debug what you cannot inspect.
§ 5 — Reliability tactics. Retries with exponential backoff, idempotent tool calls, deterministic checkpoints, the verify-then-commit pattern.
§ 6 — Security & guardrails. The threat model. Prompt injection. Excessive tool capability. The principle-of-least-privilege applied to LLM tool surfaces.
§ 7 — Evaluation. Building a real eval set. The metrics that matter (task success rate, time-to-success, cost-per-success) versus the ones that don’t (token counts, model parameter counts).
§ 8 — The pre-launch gate. The checklist before you ship. Five tests that catch most of what goes wrong.
§ 9 — Ops after launch. The cadence. Daily review of failures. Weekly eval-set updates. Monthly model-and-tool refresh. The maintenance discipline that turns a working system into a staying-working system.
Read it
The playbook lives at its own URL with a warm-paper editorial layout — signal-orange agent accent, teal verification accent throughout. Volume II of the Scaling Agentic AI series.
← Back to Autonomy