ai-registry

Use agent-browser for browser automation against real rendered web pages or web apps. Trigger when tasks require navigation, element interaction, extracting rendered data, or taking screenshots. Do not use for static fetches, direct API calls, or non-browser work.

Use when the user wants to scrape websites, crawl or map URLs, search the web, extract structured data from web pages, or run AI web agents using the Firecrawl API from the command line.

Create, revise, and validate effective skills. Use when adding a new skill or changing an existing skill's SKILL.md, bundled scripts, references, assets, scaffold templates, or validation tooling.

Must use any time a TypeScript file is read or written.

Must use whenever Bun is used.

OpenCode session storage locations and inspection guidance. Use when asked where OpenCode stores session data, chat history, local databases, or repo-specific session metadata.

Apply Go coding rules, design principles, and project conventions for maintainable Go code and repositories. Use when writing, reviewing, refactoring, or releasing Go code, or when working on Go-specific project structure, APIs, tests, CI, or binaries. Do not use for non-Go codebases or for language-agnostic tasks that do not depend on Go conventions.

User when asked to commit changes in a Git repository.

Grilling session that challenges your plan against the existing domain model, sharpens terminology, and updates documentation (CONTEXT.md, ADRs) inline as decisions crystallise. Use when user wants to stress-test a plan against their project's language and documented decisions.

Read, write, rewrite, reorganize, and archive internal .md files. Unified documentation writer that manages general internal docs, engineering designs, tickets, and AGENTS guidelines. Must be used whenever touching files in {{ env "DOCS_INTERNAL_DIR" }}.

Write, rewrite, and update `README.md` files for projects. Use when creating a project README from scratch, tightening an existing README, or updating evergreen documentation after project changes. Ground claims in the available source materials and describe the project as it exists today, not as a narration of what just changed.

Audit whether a repository's checked-in code, scripts, workflows, and docs actually define and enforce repository policies. Use when asked to verify tests, CI, pre-commit, AGENTS, deploy, or release enforcement and produce a formal policy audit.

Use whenever merging code to main/master branches.

Use whenever reviewing code.

Ask users questions via the UI. Use when you need clarification, user preferences, or confirmation before proceeding. This is the ONLY way to communicate with users - they cannot see CLI output.

Add or update content in the AI registry. Use when a user needs to create or modify skills, commands, profiles, harness overrides, vendored skills, or related registry documentation in this repository.

Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json file. Also triggers for "shadcn init", "create an app with --preset", or "switch to --preset".

Configure Codex local clients, project `.codex/` layers, profiles, providers, approvals, permissions, hooks, MCP servers, and enterprise policy files. Use when the task involves `~/.codex/config.toml`, `.codex/config.toml`, `requirements.toml`, or explaining Codex configuration precedence and override behavior.

Interact with a local Chrome-family browser session through CDP. Use only after the user explicitly approves inspecting, debugging, or interacting with a page already open in Chrome or Chromium.

Rules and guidelines for configuring the OpenCode CLI, including Agents, Commands, Skills, and Permissions.

Analyze OpenCode conversation history to identify themes and patterns in user messages. Use when asked to analyze conversations, find themes, review how a user steers agents, or extract insights from session history.

Configure Pi Coding Agent (`pi`) settings, prompts, skills, models, and repo-local `.pi` directories. Use when working with Pi config files such as `settings.json`, `models.json`, `SYSTEM.md`, `APPEND_SYSTEM.md`, `prompts/`, `skills/`, or `extensions/`, and when mapping another tool's prompts or skills into Pi's native configuration surface.

Find movie showtimes for a user-specified movie, date, time window, and geographic corridor, then return a theater-by-theater list with direct theater links and ticket links when available. Use when the user asks for nearby movie times, wants coverage across several neighboring cities, or needs browser automation because theater pages hide showtimes behind JavaScript or anti-bot flows. Do not use for reviews, plot summaries, or generic entertainment recommendations that do not require concrete showtime verification.

Guide for writing ast-grep rules for structural code search and analysis. Use when a task requires matching syntax trees, language constructs, or code patterns that depend on structure rather than plain text, such as finding specific call shapes, declarations, or AST-level relationships. Do not use for simple text searches that `rg` or filename globbing can answer without structural matching.

Use when adding Tailwind to a Bun app, fixing missing utilities in Bun-built assets, defining Tailwind v4 theme tokens with @theme, or documenting how consumers must import and scan a Bun component library.

Create and manage Gitea repositories and inspect Gitea Actions using Tea CLI and direct Gitea credential or API workflows. Use when asked to create a repo, make a "tea repo", inspect Tea logins, check whether a repo exists, configure or update a Git remote, push local branches, verify remote repository state, or diagnose CI or workflow failures on a Gitea host.

Create, inspect, edit, and publish GitHub releases with the `gh release` CLI, including writing release notes, verifying tags, managing draft/prerelease state, and uploading assets. Use when asked to create a GitHub release, fix or replace release notes, publish an existing tag, inspect release metadata, or attach release assets. Do not use for generic git tagging or repository publishing flows that do not involve GitHub Releases.

Use the gitsnip CLI to fetch full repositories or specific directories from GitHub without cloning entire repos. Trigger whenever there is a need to explore full or partial GitHub repository content, inspect remote code quickly, or pull only selected folders for analysis.

Compose, rewrite, refactor, and adapt prompts for LLM applications. Use when drafting system, developer, or user prompts; turning vague requirements into reusable prompt templates; improving prompt reliability or output format; or deciding prompt structure for grounded, long-context, tool-using, or structured-output workflows.