Three architectural-decision skills: decide-architecture (compose a software architecture stack), design-patterns (choose the right GoF / Python-idiomatic pattern), and agentic-patterns (design an LLM-agent control flow). Each branches on status — greenfield → selection interview → recommended design; refactoring → code review against the catalog → targeted improvements.
A suite of recent data-science skills: ds-star & ds-star-plus (iterative verified solving with a rubric-graded LLM-as-judge), ds-clarify (human-in-the-loop spec), ds-spike (multi-data-scientist ensemble with debate), ds-model (AIDE solution-tree + leaderboard), ds-conduct (data-aware orchestrator), ds-memory (cross-session memory), ds-verify/ds-reconcile/ds-vote/ds-search (standalone primitives), data-profile, and eda-narrative.