Resolve NixOS issues using research and sequential thinking. Use when diagnosing or fixing problems on a NixOS host -- failed builds or switches, broken services or modules, configuration errors -- or whenever the user mentions NixOS, nixos-rebuild, or…
Methodology for the user-triggered /wiggum command (do not self-invoke). An autonomous-continuation loop for long-running work -- run, checkpoint, and verify until a defined Definition of Done holds or a stop-and-escalate condition fires. Covers durable…
Build, edit, and debug Node-RED flows on John's NixOS host (vulcan). Tuned to his actual plugin set, wiring conventions, naming style, and to the nodered_events PostgreSQL log + Grafana dashboard for chain tracing. Use whenever the user mentions Node-RED,…
Offload safe, independent subtasks to concurrent subagents while continuing your own work as the coordinator, then integrate what they return. Use when mid-task and some work could run in parallel without conflicting -- research, generating a standalone new…
Use when the user requests a thorough pre-PR code review of the current branch — e.g. "run a validated code review". "validated" is a key word
Fix all issues — no exceptions, no excuses. Fix every finding uncovered during the work, here and now. "Out of scope," "pre-existing," and "follow-up ticket" are not acceptable framings. Fixes go upstream, everything changed gets a real test, and no reward…
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Architecture-alignment code review that detects when a change evades the codebase's existing abstractions instead of extending or correcting them: parallel implementations built beside existing ones, special-case dispatch on identity, semantics carried by…