agent-tooling

Design or repair one Apple App Intents system integration while preserving app-owned data, concurrency, routing, privacy, and availability boundaries. Use for AppIntent and AppEntity work that exposes an action or entity through Siri and Shortcuts; Spotlight; widgets or controls; snippets; the Action button; App Schemas; onscreen context; or AppIntentsTesting. Do not use for ordinary in-app SwiftUI work or broad Apple-platform setup that does not cross an App Intents surface.

Diagnose and tune one user-visible performance problem in a Swift, SwiftUI, React Native, or Expo mobile feature by tracing the first missed deadline across motion, rendering, state publication, media, storage, network, and observability, then applying and verifying the smallest causal change. Use when an interaction, animation, scroll, keyboard transition, first paint, or loading path is jittery, stalled, late, or less seamless than expected and the bottleneck is uncertain. Do not use for broad app performance audits, motion ideation or craft review without a runtime symptom, recording-only analysis, backend-only optimization, or release hardening.

Use when creating or initializing a new software repository, or turning an empty or newly generated project directory into a maintained local and GitHub-ready repository. Applies across Swift, Expo or React Native, Python, React or TypeScript, and other stacks. Do not use for auditing or modernizing an established repository, adding one CI job, implementing product features, deploying, or publishing.

Use when a user wants to brainstorm or compare multiple runnable UI, interaction, layout, visual, or motion directions in an existing web or native app before choosing one. Do not use for exact implementation of a selected design, production delivery, audits, or requested static or generated mockups.

Take a substantial production feature from an idea or selected behavior through specification, coordinated implementation, integration, and staged user availability. Use when delivery spans meaningful product or architecture decisions, multiple repositories or subsystems, important data or security boundaries, client and backend compatibility, or controlled rollout. Do not use for UI direction exploration, bounded feature spikes, hardening-only work, deployment of an already completed build, small fixes, or ordinary isolated changes.

Build the smallest decision-producing slice that validates a feature's value or technical feasibility before production investment. Use when a user asks for a feature spike, proof of concept, capability probe, or bounded experiment to decide whether or how to proceed. Do not use for comparing multiple UI directions, delivering an already-approved production feature, general research that needs no runnable evidence, production hardening, or experiments inside an established implementation.

Review one completed UI component or bounded surface against the design system evidenced in its repository, classify accidental deviations, intentional exceptions, system gaps, and uncertainty, then apply only approved corrections. Use for post-build UI conformance or normalization. Do not use for UI ideation, repository-wide audits, exact design translation, or general production hardening.

Use when a user supplies a UI screen recording or video and asks to dissect, diagnose, reverse engineer, compare expected and observed behavior, or explain how it could map to a target repository. Do not use for static screenshots, code-only motion review, video editing, or implementation requests without video evidence.

Audit one or more code repositories for design-system drift, hardcoded visual decisions, inconsistent token or primitive use, overlapping UI components, missing system capabilities, and enforceable rules. Produce a prioritized read-only health and consolidation report grounded in source and rendered evidence. Do not use for aligning one component, UI ideation, greenfield system creation, Figma-only audits, or automatic refactoring.

Reviews a Mintlify documentation site's information architecture, editorial quality, source fidelity, and launch readiness in navigation order, then makes approved targeted edits. Use for Mintlify site audits, content passes, or page-level reviews; do not use for general code review, changelog entries, or generated reference pages.

Writes or audits reader-facing entries in a Mintlify changelog.mdx, including tags, headings, voice, length, and removal of implementation trivia. Use for Mintlify changelog entries and content reviews; do not use for repository CHANGELOG.md files, release-note generation, or git-log summaries.

Generates drift-checked CLI, MCP, or API reference documentation for a Mintlify site from authoritative sources and wires regeneration checks. Use when adding or repairing source-backed Mintlify reference pages; do not use for hand-authored guides, source-code docstrings, or non-documentation code generation.

Scaffolds a new Mintlify documentation site with a project-specific information architecture, docs.json, starter pages, assets, and local documentation toolchain. Use when creating a new Mintlify site or docs-mintlify tree; do not overwrite an existing site, author finished product content, deploy, or mutate external hosting without explicit approval.