Kimi Code Tools Guidebook

Test-driven development workflow (red-green-refactor). Use when building features or fixing bugs test-first, or when you want integration tests written before the implementation.

Reviews changes since a commit, branch, or merge-base along two axes: whether the code follows the repo's documented standards, and whether it matches what the originating issue or PRD asked for.

A structured diagnosis loop for hard bugs and performance regressions. Use when something is broken, throwing, failing, or slow and the cause isn't obvious.

Interactive QA session: report bugs conversationally and the agent explores the codebase in the background, then files well-formed GitHub issues.

Investigates a question against high-trust primary sources and captures findings as a Markdown file in the repo. Good for delegating reading legwork to a background agent.

Builds a throwaway prototype to answer a design question — sanity-check whether a state model feels right or explore what a UI should look like before committing to real code.

Creates a detailed refactor plan with tiny, safe commits via a structured interview, then files it as a GitHub issue. Use before any large restructuring.

Guides resolution of an in-progress git merge or rebase conflict.

Migrates test files from TypeScript 'as' type assertions to @total-typescript/shoehorn partial test data, making tests less brittle.

Creates exercise directory structures with sections, problems, solutions, and explainers that pass linting — for course or workshop content.

Sets up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests so bad code never gets committed.

Adds hooks that block dangerous git commands (push, reset --hard, clean, branch -D) before they execute. A safety net against destructive operations.

Grills you relentlessly about a plan, decision, or idea to stress-test the thinking before you build on it.

Shared vocabulary for designing deep modules: improving a module's interface, finding deepening opportunities, deciding where a seam goes, making code more testable and AI-navigable.

"Design it twice" as a service: spins up parallel sub-agents to generate radically different interface designs for a module so you can compare shapes before choosing.

Builds and sharpens the project's domain model — pins down terminology and a ubiquitous language, records architectural decisions.

Generates images and video via Higgsfield AI — GPT Image 2 for images/design/text, Seedance 2.0 for video, Marketing Studio for ads with avatars and hooks.

Brand-quality product imagery: studio shots, lifestyle images, Pinterest-style visuals through Higgsfield's product-photoshoot prompt enhancement.

Marketplace listing assets: compliant main images, secondary product images, and A+ style content modules for product detail pages.

Trains a "Soul Character" — a personalized model on a person's face for identity-faithful image and video generation (digital twin / avatar work).

Searches, creates, and organizes notes in your Obsidian vault with wikilinks and index notes.

The skill that found this list — searches the open agent-skills ecosystem and installs new skills via npx skills.

Answers questions about the Kimi Code product itself — CLI usage, configuration, slash commands, features, membership/quota, error codes — from official docs.

Inspects or edits Kimi Code's own config: config.toml (model, provider, permissions, hooks) and tui.toml (theme, editor, notifications).

Helps craft a well-specified /goal objective for goal mode — turns a rough intention into a completion contract with a clear finish line and stop rule.

React and Next.js performance and pattern guidance from Vercel Engineering. The most-installed skill in the ecosystem (593K installs) and a direct fit for the app/ code in this project.

Install: npx skills add vercel-labs/agent-skills@vercel-react-best-practices

Official Anthropic document-processing skills (100K+ installs each). Directly relevant to this repo's deck and guidebook pipeline (scripts/build_deck.py, build_guidebook.py → PPTX/PDF).

Install: npx skills add anthropics/skills@pptx · npx skills add anthropics/skills@pdf

Design system architecture guidance (12.4K installs) — tokens, component APIs, and patterns. Fits the SBA Design System work directly.

Install: npx skills add wshobson/agents@design-system-patterns

OpenAI's official PDF skill (10.8K installs). A solid alternative to the Anthropic version for PDF reading and generation.

Install: npx skills add openai/skills@pdf

Tailwind-based design system guidance (56.9K installs). Only worth adding if the project moves off CSS modules + tokens to Tailwind.

Install: npx skills add wshobson/agents@tailwind-design-system