Skill Library

Agent delegation patterns — Single, Sequential, and Parallel sub-agents. Includes rules and review pass limits.

Design REST or GraphQL APIs, review API specifications, or establish API design standards. Does not cover API gateway configuration, authentication implementation, or deployment. Produces well-structured, versioned APIs with consistent naming, error handling, and pagination.

Implement Clerk authentication integration patterns including user invitations, invitation lifecycle, and domain error classes. Use when integrating Clerk auth into Next.js backends, handling invitation flows, or writing tests for Clerk-protected endpoints.

Shared function to create a fully enriched Linear issue — wraps create-issue.sh + validate-issue + complexity label + DoR validation + Prep Summary comment. Called by /new, /go (new findings), and /approve (sub-issues).

Generate diagrams from text descriptions using AI image generation. Produces hand-drawn Excalidraw-style visuals for pipeline flows, architecture overviews, org structures, and spatial relationships. Multi-turn refinement: generate, review, revise until approved, then place alongside the document. Requires GEMINI_API_KEY environment variable.

Performs security-focused differential review of code changes (PRs, commits, diffs). Adapts analysis depth to codebase size, uses git history for context, calculates blast radius, checks test coverage, and generates comprehensive markdown reports. Automatically detects and prevents security regressions.

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Generate working Next.js prototypes with shadcn/ui from text descriptions using Gemini image generation. Does not cover production deployment, backend API implementation, or database setup. Produces a design mockup image plus fully functional React components ready to run with `npm run dev`.

Bootstrap a new child project from the prototype scaffold. Produces a fully configured repo with GitHub remote, Linear integration, CLAUDE.md, MCP config, and passing quality gates.

Next.js best practices - file conventions, RSC boundaries, data patterns, async APIs, metadata, error handling, route handlers, image/font optimization, bundling

Activate for any direct PDF action a user wants executed right now: generate a PDF from a document file (markdown, HTML, RST, Word), print a file to a printer, preview a document's formatted appearance, email a file as a PDF attachment, merge multiple PDFs, split a PDF into pages, extract specific pages, rotate pages, add watermarks, OCR a scanned PDF, extract text or tables, fill a form, or password-protect a file. Has the pandoc commands, Typst template, and Resend API integration needed — load before starting, not after. Skip for: reading or summarizing an existing PDF, writing PDF-generation code, setting up CI pipelines, or comparing PDF libraries.

Push a feature branch to Vercel, test the preview deployment, create a PR, merge, and verify the main preview. Executed by the orchestrator directly (not a sub-agent).

Atomic function to process media URLs (video, image, Linear attachment) and return structured output. Wraps the video skill's core processing. Called by /new and /go before DESIGN.

Locked technology decisions, critical rules, quality gates, and stack reference for all framework child projects

High-stakes prompt engineering methodology for real-world decisions and deep research. Generates structured, evidence-driven prompts using the PROMPT-74 framework: Role + Constraints + Source of Truth → Evidence-driven analysis → Actionable deliverable. Use when the user needs a thorough, professional-grade prompt for complex decisions, deep research, strategy analysis, or any task requiring expert-level output with verification. Trigger phrases: "prompt74 this", "prompt74 deep research for", "high-stakes prompt", "generate a decision prompt", "ownership prompt".

Build transactional email templates with React Email and send via Resend. Does not cover marketing email campaigns, email list management, or SMTP server configuration. Produces type-safe, reusable email templates with shared layouts, preview server support, and Resend delivery integration.

Sentry error tracking workflow — setup, source maps, session replay, release management, and alert configuration. Use when integrating Sentry into a project, debugging production errors, or configuring monitoring. Sentry is a Locked Decision for error tracking.

Build UI with shadcn/ui in framework projects using unified radix-ui imports and Tailwind v4. Does not cover general React patterns, backend logic, or non-shadcn component libraries. Produces correctly themed components with oklch colors, proper Radix imports, and React Hook Form + Zod integration.

Create new skills from scratch, improve existing skills, run evals to measure performance, or optimize skill descriptions for triggering accuracy. Does not cover skill installation, plugin packaging, or skill marketplace publishing. Produces a tested SKILL.md with eval results, benchmarks, and an optimized description.

Postgres performance optimization and best practices from Supabase. Use this skill when writing, reviewing, or optimizing Postgres queries, schema designs, or database configurations.

Supabase project workflow — auth, storage, edge functions, migrations, RLS, and MCP integration. Use when setting up Supabase in a project, writing migrations, configuring auth, or working with edge functions. Complements supabase-postgres-best-practices (query/schema optimization) with project-level workflow patterns.

Investigate any bug, test failure, or unexpected behavior using a root-cause-first methodology. Does not cover performance profiling, load testing, or monitoring setup. Produces a verified root cause diagnosis with a single targeted fix backed by a failing test.

Plan, build, review, or improve UI/UX for web and mobile applications. Does not cover backend logic, API design, or database schema. Produces bold, intentional design systems with style recommendations, color palettes, typography pairings, and implementation-ready code across 9 technology stacks.

Shared function to validate and normalize a Linear issue — type label inference, estimate defaulting, complexity derivation, description structure validation, monochrome prefix application, and assignee/project checks.

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

Process a video URL — transcribe, summarize, and output to terminal, Linear, or Obsidian with executive summary, chapters, and references.

Web tool routing, detection, and fallback chains for any skill or agent that accesses websites. Provides credit-aware tool selection — chooses the cheapest working tool first, escalates only on failure. Use this skill when building or running any research workflow that reads web pages or searches the web. Skills invoke this directly; agents (which can't load skills at runtime) reference this as the canonical source and carry quality-optimized overrides inline. Other skills (competitor-research) invoke this for shared routing logic and add their own domain-specific overrides.

Toolkit for interacting with and testing local web applications using the Playwright MCP server. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

Use when you have a spec or requirements for a multi-step task, before touching code