rakaadi/agent-kit

Guide for editing existing and writing inline comments and JSDoc. Use when generating comments for documenting new components or functions, non-obvious code behaviour, or complex logic.

Execute or resume an attached or referenced implementation plan through repository inspection, implementation, verification, sequential spec-compliance and code-quality review, and human-authorized architectural review when warranted. Use for full or partial plan execution, including free-form instructions to select phases or task IDs, skip work, delegate a task independently, reserve checks for a human, or impose task-specific constraints.

Close repository diagnostic findings from TypeScript, ESLint, and React Doctor. Use when asked to verify or fix project-wide tsc errors, ESLint violations, React Doctor findings, or a scoped batch from any of those tools.

Stop. That last message did not land, restate the last message in plain human language, with no jargon.

Design the code shape before implementation for substantial features or refactors that introduce or reshape modules, interfaces, types, file layout, or control flow. Use before writing code when meaningful structural choices remain; skip small fixes and mechanical edits.

Assess flagged code architecturally and recommend justified retention or structural improvement, with approval before changes.

Dispatch subagents for bounded independent work. Use whenever Codex is about to delegate or spawn a subagent, or when the user asks for parallel agents, context isolation, specialist work, or an independent review. Decide whether delegation helps, select an available agent, write an outcome-first contract, schedule work safely, and integrate every result.

Use when you have a spec or requirements for a multi-step task, before touching code. Supports either Markdown plans or review-oriented HTML plan artifacts.

Guide for writing a unit test. Use when writing a unit test for functions or components, and when fixing a bug in existing code.