majestic-engineer

Review a concrete change set for correctness, regressions, security, data risk, unnecessary complexity, and missing tests, then give evidence-backed findings or a release verdict. Use when reviewing a diff, pull request, patch, staged changes, or deciding whether code is ready to merge or ship.

Simplify settled, recently changed code for reuse, quality, and efficiency while preserving behavior. Use for explicit code simplification requests after implementation and before handoff.

Review code for algorithmic complexity and performance hotspots without changing behavior. Use when auditing inefficient loops, repeated scans, rendering work, N+1 queries, or proposed performance optimizations.

Create an implementation plan grounded in the existing repository, with explicit scope, ordered changes, risks, and verification. Use when a feature, fix, migration, or refactor needs an executable technical plan before code changes begin.

Apply the MSW deletion rule to keep plans, implementation, tests, remediation, reviews, artifacts, and process limited to work necessary to satisfy or prove the requested outcome. Use when the user invokes MSW or the deletion rule, when work risks expanding beyond scope, or before adding code, tests, files, abstractions, limits, review findings, or follow-up work.

Design persistent multi-agent systems with explicit roles, ownership, handoffs, state, permissions, context budgets, and failure handling. Use when building an agent product or durable workflow whose work genuinely requires multiple specialized agents.

Review implementation plans, PRDs, brainstorm handoffs, and feature specifications as whole delivery artifacts for implementation readiness. Use when flows, sequencing, dependencies, scope, operational risk, or verification should be challenged before handoff. Not for a dedicated requirement-statement or requirements-catalogue quality audit.

Design structured application logging for observability and incident debugging. Use when adding JSON logs, correlation or trace identifiers, context propagation, log-level policy, redaction, or production debugging signals.

Review automated tests for behavioral coverage, assertion quality, isolation, and missing failure cases. Use when assessing whether a change is sufficiently tested or whether a test suite gives reliable shipping confidence.

Convert an approved plan, specification, or settled conversation into the smallest dependency-aware set of vertical implementation tasks, then optionally create them in an authorized task system or as local Markdown. Use when work spans multiple implementation sessions, needs explicit blockers or safe parallel execution, or requires expand-migrate-contract sequencing for a wide refactor.