AI Engineering Patterns

Browse from the Patterns menu — twelve operational envelopes for consequential AI-assisted work.
Version: 1.2 | Updated: 2026-07-16

---

Purpose

Patterns are runnable specifications, not essays and not a sidebar dump of thirteen stub files. Open a pattern README, follow its Mermaid workflow, use its prompt, and keep humans at the approval gates.

Catalog

Pattern	Outcome	Start here
Planner	Decision-complete execution contracts	Multi-file / cross-boundary work
Reviewer	Severity-calibrated findings	Before merge
Context	Minimal high-signal packages	Before planning or coding
Debug	Hypothesis-driven isolation	Defects with repro
Migration	Safe transitions with return path	Data/service moves
Legacy Modernization	Incremental strangler seams	Behavior-preserving rewrites
Architecture	Scenario-governed design	Material topology choices
Optimization	Profile-guided improvement	Measured bottlenecks
Enterprise	Governed multi-tenant AI	Policy + jurisdiction
Release	Progressive exposure + abort	Production promote
Hotfix	Smallest reversible containment	Active incident
Root Cause	Evidence-bound learning	After containment

How to apply

1. Match the problem to a pattern trigger (read When / Do not).
2. Assemble context (handbook).
3. Follow workflow.md — do not skip human gates.
4. Use prompt.md variables; do not rewrite the contract casually.
5. Run hooks + checklist before calling done.

Component files (problem.md, hooks.md, …) support the README — they are not meant to flood navigation. Use in-page links from each pattern README.

Practitioner handbook

For daily practice (skills vs prompts, HITL, PR review, tokens): ../handbook/README.md.

Validate

node content/patterns/validate-pattern-library.mjs

Changelog

• 2026-07-16: Restored catalog landing for mega-menu browse; pointed to handbook; demphasized stub-file navigation.