Agent Claude Code
Designs implementation plans for large or ambiguous work — new subsystems, schema changes, cross-cutting refactors. Use before starting anything that spans multiple subsystems or changes the shared event schema.
Agent Claude Code
Full-repository technical audit agent. Use when the user asks for a codebase audit, health check, tech-debt review, or a prioritized improvement plan. Produces a Repo Map, severity-rated Audit Report, Improvement Strategy, and milestone-based Task Plan. Read-only — never modifies code.
Agent Claude Code
Reviews diffs and modules for correctness bugs, race conditions, and cross-process issues. Use after implementing a feature, before committing, or when something behaves oddly across main/renderer/hook boundaries.
Agent Claude Code
Read-only codebase search and explanation agent. Use for "where is X handled", "how does Y flow work", tracing event paths across bridge/installer/pollers, and summarizing subsystems before making changes.
Agent Claude Code
Security review of changes and subsystems. Use for anything touching the HTTP bridge, hook installers, credential/token handling, webhook senders, guardrails, or the updater.
Agent Claude Code
Writes and fixes Vitest tests. Use for adding coverage to bridge normalization, hook installers, guardrails, the Zustand store, or bubble animation variants, and for fixing failing tests.
Agent Claude Code
Implements renderer UI work — bubbles, settings tabs, animations, layout. Use for scoped React 19 + Tailwind 4 + Framer Motion changes in src/renderer/.
Agent Claude Code
Senior product-designer + frontend-engineer UI/UX audit agent. Use when the user wants a deep UX review, design critique, screen-by-screen sweep, motion/animation spec, accessibility check, or a prioritized "what to fix" list for the renderer. Produces docs/UXAUDIT.md. Read-only — never modifies code.