Marcus Skills

Build programmatic videos with Remotion 4. Covers compositions, components, audio sync, rendering, and deployment. Use when creating video content with React/TypeScript.

Use Braid's CLI to operate the daily podcast pipeline, generate images with local models, and inspect distributed inference. Use when checking whether the daily run succeeded, inspecting or recovering a workflow run, running preflight, rebuilding or deploying the site or an episode, opening the Braid dashboard, generating images with local adapters, or inspecting Braid workers, queues, claims, source scores, and dead-letter jobs.

Create and maintain DESIGN.md files — a design system specification optimized for AI consumption. Use when creating a project's design system file, extracting tokens from existing CSS/code, auditing design consistency, or onboarding an agent to a project's visual language.

Use Roc icon library (@marcus/roc) in projects. Covers React, Svelte, and HTML sprite usage, all 4 styles, and migration from other icon sets.

Orchestrate development work through sub-agents using td for state. Use when given a td task ID, text idea, markdown plan, or td epic to execute through plan-implement-review loops, or when asked to triage the pending review queue.

Write and update durable project, implementation, migration, and architecture plans. Use when creating a new plan document, revising an existing plan, choosing where a plan belongs in a repository, or splitting a large plan into an indexed set of documents.

Create reproducible TUI screenshots and GIF demos for documentation with Betamax. Use when asked to capture terminal screenshots, record demos, or author .keys files.

Present technical information, decisions, tradeoffs, and code in chat using Google Developer Documentation Style Guide standards. Use when answering technical questions, presenting architecture/design choices, explaining technical concepts, structuring procedural steps, or formatting technical communication in chat.

Write naturally and avoid AI-detectable patterns. Use when (1) generating any written content, (2) reviewing/editing text for AI-like patterns, (3) user asks to make writing sound more human/natural, or (4) improving text that sounds robotic or generic. Covers vocabulary, structure, tone, and formatting tells that signal AI authorship.

Use the Perch plan CLI to create, read, update, and manage architectural plans. Covers project lifecycle, tree building, reviews, snapshots, facets, inputs, and export. Use when asked to retrieve a plan by name, build a plan tree, run a review, or interact with the planner in any way.

Design, implement, refactor, and review software that exposes owned capabilities across human interfaces, agent interfaces, and APIs without duplicating business logic or allowing surfaces to drift. Use when choosing or evolving UI, TUI, native, CLI, API, or MCP surfaces; adding a feature to a multi-surface product; preparing a clean future API boundary; reviewing surface-only behavior; or designing parity and contract tests. Keep selected surfaces semantically aligned through one shared application core while allowing each surface appropriate ergonomics and transport behavior.

Search and query a Lightroom CC photo catalog via a SQLite+FTS5 index. Use when the user wants to find photos, browse their library, get stats, search by keyword/rating/camera/date, or work with Lightroom catalog data. Covers catalog location, indexing, search queries, and the index database schema.

Guide for programmatically interacting with Claude Code's task/todo system from external code. Use this skill whenever the user wants to: monitor or track Claude Code tasks from scripts, build automations that react to todo updates, create hooks that enforce task completion criteria, build custom task management MCP tools, use the Agent SDK to watch or control task progress, stream task updates via the CLI, integrate Claude Code's task system with external services, or orchestrate work across subagents. Also trigger when users mention "TodoWrite", "TaskCompleted", "Agent SDK tasks", "programmatic tasks", "task hooks", or "claude code automation" — even if they don't explicitly say "skill".

Index and search Adobe Lightroom CC catalogs on macOS. Build a searchable SQLite+FTS5 database from the .mcat catalog, with incremental sync, full-text search, EXIF/keyword filtering, and stats. Use when user wants to query their Lightroom library, find photos by metadata, or set up automated catalog indexing.

Create and debug rich link previews (Open Graph + Twitter/X Cards) for iMessage, WhatsApp, Slack/Discord, and social platforms. Use when user asks about og tags, twitter cards, link previews, social sharing metadata, or when a shared link preview looks wrong or missing.

Generate and render technical diagrams (C4, sequence, flowchart, architecture, ER, state, infrastructure) using Mermaid.js for inline web rendering or D2 for static PNG/SVG export with hand-drawn sketch style. Covers syntax for all diagram types, browser rendering, dark theme integration, LLM generation best practices, and accessibility.

Generate images via OpenAI (gpt-image-1 / gpt-image-1.5), Google Gemini (Imagen 4), or AWS Bedrock (Amazon Titan Image Generator v2). Includes composite diagram system (Mermaid + Titan backgrounds + ImageMagick) for slide-quality diagrams with perfect text. Use when asked to generate images, create illustrations, compose diagrams, build slide graphics, or use Bedrock/Titan for image generation.

Conduct architectural reviews for new or existing software projects. Use when: (1) starting a new project and need architectural foundations, (2) reviewing architecture decisions before implementation, (3) creating ADRs or design docs, (4) evaluating technology selections, (5) assessing non-functional requirements, (6) setting up project scaffolding with built-in quality. Combines enterprise ARB governance with modern agentic coding patterns.

Extract, decode, and query iMessage conversations from macOS chat.db. Use when user needs to access iMessage history, export conversations, search messages, decode attributedBody fields, convert Apple timestamps, transcribe voice messages, or create clean SQLite databases from iMessage data. Handles NSKeyedArchiver decoding, artifact cleanup, schema navigation, contact name resolution via AddressBook databases, and voice message transcription via yap.

Create premium software demo and product videos that convert. Covers every phase from thumbnail and script through production, narration, transitions, sound design, and distribution. Use when planning, scripting, storyboarding, or reviewing any software marketing video — explainers, product tours, feature highlights, launch announcements, customer stories, or comparison videos.

Comprehensive guide to Claude Code's local data storage, session formats, token tracking, and metadata structures. Use when building dashboards, analytics tools, migration scripts, or any application that needs to read Claude Code's local data. Covers JSONL session format, stats-cache.json structure, settings hierarchy, file history, and all ~/.claude directory contents.

Use when the user asks for help designing (or redesigning) a website/landing page/SaaS UI/portfolio. Delivers UX, typography, layout, tokens, components, interaction guidance, and production code—while avoiding generic AI design. Provides both strategic design direction AND implementation.

Implement single-writer queue pattern for SQLite in multi-process environments. Use when building CLI tools, local apps, or agent systems where multiple processes may write to SQLite concurrently. Covers file locking (flock/LockFileEx), WAL mode, timeout handling, and diagnostic error messages. Works for Go, Rust, Python, JavaScript/Node.

Plan, build, refactor, and review software from the user journey outward, using a working steel thread for new capabilities and tracing the affected journey for existing behavior. Use by default for software planning, implementation, debugging, refactoring, testing, technical decision-making, and code review unless the user explicitly requests enterprise architecture governance or invokes a more specialized conflicting approach. Optimize for useful behavior, iterative delivery, clear conventions, replaceable seams, and a maintainable agent and developer experience without speculative hardening or overarchitecture.

Create state-of-the-art interactive online courseware with AI agents. Covers course architecture, multimedia generation (narration, video, images, sound), interactive patterns (branching scenarios, simulations, gamification), SCORM/xAPI/cmi5 packaging for LMS delivery, and collaborative agent-human workflows. Use when building courses, training content, e-learning modules, or educational experiences.

Linear-inspired design system patterns for building keyboard-first, high-density, dark-first web applications. Use when (1) building a new design system or design tokens inspired by Linear's aesthetic, (2) implementing keyboard-first navigation with command palettes and vim-style shortcuts, (3) designing admin tools, developer tools, or SaaS UIs that prioritize speed and information density, (4) choosing color systems, typography, animation, or feedback patterns for professional/engineering-focused apps, (5) user mentions Linear, Linear-style, or asks for a clean/minimal/keyboard-first design approach.