adaptive-engineering

Select and execute the least process necessary for a software-engineering task using FAST, STANDARD, or DEEP mode, with explicit reuse, complexity, simplification, and verification gates. Use for implementation, debugging, refactoring, code review, architecture, CI/CD, infrastructure, security, data, and public-contract work when process depth should scale with risk.

Review a change for correctness, regressions, security, compatibility, operational risk, and unnecessary complexity, prioritizing actionable findings over style noise.

Inspect the repository for existing behavior, conventions, native capabilities, and installed dependencies before creating new code or abstractions.

Produce a right-sized implementation plan with acceptance criteria, complexity budget, verification, and rollback proportional to the selected mode.

Review an implementation or diff for avoidable complexity, unnecessary code, dependencies, abstractions, configuration, and duplication before completion.

Debug failures by reproducing, narrowing, forming evidence-based hypotheses, and verifying the root cause before applying the smallest durable fix.

Classify a software task as FAST, STANDARD, or DEEP based on risk, scope, ambiguity, reversibility, and blast radius. Use before non-trivial implementation or when choosing process depth.

Establish concrete evidence that a change works by selecting and running the strongest practical tests, checks, builds, or integration validations for the task.

Implement a planned change in small coherent increments, preserving repository conventions and verifying behavior as work progresses.