Find deepening opportunities in the codebase — refactors that reduce coupling, simplify interfaces, and improve AI-navigability. Use when improving architecture, finding refactoring opportunities, or making a codebase easier to work with.
Use before any feature work — explores intent, asks clarifying questions one at a time, produces a design doc. Invoke before writing any code or plan.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Use when user says "caveman mode", "talk like caveman", "less tokens", or invokes /caveman.
Disciplined diagnosis loop for hard bugs and performance regressions. Reproduce → minimise → hypothesise → instrument → fix → review. Use when something is broken, throwing errors, or regressing.
Interview the user or a design relentlessly until reaching shared understanding. Walk down every branch of the decision tree, resolving dependencies one by one. Use when stress-testing a plan, clarifying design, or the user says "grill me".
Use at the start of any session to understand which skill to invoke. Maps the full workflow — feature development, bug fixing, architecture, and utilities — for a C++ game server project with SVN.
Compact the current conversation into a handoff document for another agent or session to pick up.
Disciplined implement→build→review→commit loop for C++ game server features. Use when executing a plan task, fixing a bug, or making any code change. One vertical slice at a time.