Zlabs Coding Guidelines Plugin
Zlabs coding guidelines packaged as agent skills for TypeScript, React/Next.js, NestJS, tRPC, Prisma, Better Auth and Turborepo codebases.
What It Does
The agent (Hrida, Claude Code or Codex) sees each skill's name and description and loads the full guideline only when the task matches — e.g. writing a React component pulls in vercel-react-best-practices and no-use-effect; touching a NestJS module pulls in nestjs-best-practices. You can also invoke any skill directly with /<skill-name>.
| Skill | Use it for |
|---|---|
zero-tech-debt | Rework a change as if the intended design existed from day one; delete compatibility cruft |
pit-of-success | Make the correct behavior the default path for APIs, hooks, helpers and props |
code-refactor-review | Review diffs for reuse, composition, consistency and slop |
no-use-effect | Replace useEffect with derived state, event handlers, data-fetching libraries or useMountEffect |
typescript-advanced-types | Generics, conditional/mapped/template-literal types, utility types |
vercel-react-best-practices | React/Next.js performance: data fetching, bundles, rendering |
vercel-composition-patterns | React composition patterns that scale; avoiding boolean-prop explosion |
nestjs-best-practices | NestJS modules, DI, security and performance |
nestjs-trpc | tRPC routers and procedures inside NestJS |
better-auth-best-practices | Better Auth server/client setup, adapters, sessions, plugins |
prisma-database-setup | Prisma with PostgreSQL, MySQL, SQLite, MongoDB and others |
turborepo | Turborepo pipelines, caching and monorepo structure |
Installation
The same skills/ folder works in three agents. Each reads its own manifest:
| Agent | Plugin manifest | Marketplace file (repo root) |
|---|---|---|
| hrida-agent-sdk | .hrida-plugin/plugin.json | .hrida-plugin/marketplace.json |
| Claude Code | .claude-plugin/plugin.json | .claude-plugin/marketplace.json |
| Codex | plugin.json (and legacy .codex-plugin/plugin.json) | .agents/plugins/marketplace.json |
hrida-agent-sdk
hrida-agent-sdk plugin marketplace add Hrida-AI/hrida-plugins-official
hrida-agent-sdk plugin install zlabs-coding-guidelines@hrida-plugins-official --scope project
Requires a hrida-agent-sdk build that recognizes .hrida-plugin/ manifest directories.
Claude Code
claude plugin marketplace add Hrida-AI/hrida-plugins-official
claude plugin install zlabs-coding-guidelines@hrida-plugins-official --scope project
Or inside a session: /plugin marketplace add Hrida-AI/hrida-plugins-official, then /plugin install zlabs-coding-guidelines@hrida-plugins-official.
Codex
codex plugin marketplace add Hrida-AI/hrida-plugins-official
Then open /plugins in Codex and install Zlabs Coding Guidelines.
Codex also reads plain skills from .agents/skills/ in any repo, so copying the skills/* folders into a repo's .agents/skills/ works without the plugin.
For any of the three, a local checkout path works in place of Hrida-AI/hrida-plugins-official.
--scope project records the plugin in the repo's settings so everyone on the team gets it; use --scope user to enable it for all your projects instead.
Keeping manifests in sync
name, version and description are repeated in .hrida-plugin/plugin.json, .claude-plugin/plugin.json, .codex-plugin/plugin.json and plugin.json, and the entry appears in all three marketplace files. Bump them together.
Always-on rules
Skills are loaded on demand. Rules that must apply to every change (naming, forbidden patterns, required checks) belong in the repository's project instruction file as well, with a pointer to the relevant skill for detail.
Updating the skills
The skills are copied from zlabs-hrida/.agents/skills. To refresh, re-copy the skill folders into skills/, bump version in every manifest (see above), and run the frontmatter validator:
bun .github/scripts/validate-frontmatter.ts plugins/zlabs-coding-guidelines
See THIRD_PARTY_NOTICES.md for the origin and license of each third-party skill.
Author
Zlabs Innovation