JayZeeDesign/claude-code-vault

JayZeeDesign/claude-code-vault

Agent Claude Code

Use to run test suites, typechecks, and lints, and report results structurally. Offloads noisy test output from the orchestrator's window: a failing suite can dump 30K tokens of stack traces; this agent returns the 20 lines that matter.

JayZeeDesign/claude-code-vault

Agent Claude Code

Use for research tasks that would flood the main session's context: comparing libraries or services, digging through docs, investigating how an unfamiliar API works, or answering "what are the options for X". Read-only by design; returns a compact summary, never code changes.

JayZeeDesign/claude-code-vault

Agent Claude Code

Use to review a diff, branch, or recent commits before merge. Deliberately receives zero conversation context: fresh eyes cannot share the implementer's biases. Give it only the diff location and acceptance criteria, never the reasoning behind the implementation.

JayZeeDesign/claude-code-vault

Agent Claude Code

Use when building or modifying user interfaces with shadcn/ui components and blocks: new UI components, design changes, or complete UI features. Specializes in leveraging shadcn's component library and block patterns for fast, consistent interface work.

JayZeeDesign/claude-code-vault

Agent Claude Code

Use for Vercel AI SDK v5 implementation work: chat interfaces, tool calling, streaming, agentic loop control, type-safe custom messages, or migrating from v4. Covers useChat, streamText, generateText, tools, and the React/Vue/Svelte/Angular integrations.