using-skillslisted

<SUBAGENT-STOP> If you were dispatched as a subagent for one specific task, skip this skill and do that task. </SUBAGENT-STOP> ## The rule Before you respond or act, including before clarifying questions or exploring code, check the available skills. If one plausibly applies, load it with the Skill tool and follow it. Say which one: "Using <skill> to <purpose>". If it turns out not to fit, drop it and say so. Precedence: the user's direct instructions, then project files (CLAUDE.md, AGENTS.md), then skills, then your defaults. Process skills come before platform and domain skills. The process skill sets the approach; platform and domain skills supply the specifics. ## Size the process to the work | Kind | Looks like | Path | |---|---|---| | Spike | "can we…", a feasibility question, throwaway code | State the question and the probe, get a nod, investigate, report a recommendation | | Bounded | A small change to a flow that already exists in this repo | Short design in chat → approval → test-driven-development → verification-before-completion | | Architectural | A new project, subsystem or interface, or a change across components | brainstorming → spec → writing-plans → executing-plans → requesting-code-review | When unsure, take the heavier path. If hidden complexity shows up mid-task, stop, say so, and move up a path. Never move down mid-task. ## Where to start | Situation | Skill | |---|---| | Build, add or change behavior | brainstorming | | Bug, failing test, unex