AI Engineering

Brainstorming helps turn an idea into a design and spec before any code is written. Use it to clarify requirements, explore approaches, and get approval before implementation.

Cloudflare Agents SDK builds stateful AI agents on Cloudflare Workers with durable workflows, callable RPC, scheduling, MCP, email, and streaming chat. Use it when creating agent-backed apps or worker-based automation.

Deep Research breaks a question into sub-questions, gathers sources, cross-checks claims, and synthesizes a cited report. Use it for topics that need depth, sourcing, and conflict resolution.

Deep Research Agent decomposes a question, gathers and verifies information from many sources, and synthesizes a structured report. Use it for deep dives, competitive analysis, market research, academic surveys, or any task needing thorough source-backed investigation.

Effect helps you work with Effect v4 and effect-smol TypeScript code in this repo. Use it when implementing services, schemas, workflows, and tests that follow the project's local patterns.

Improve surveys a codebase and produces prioritized, self-contained implementation plans for another agent to execute. It is used for audits, improvement opportunities, and handoff plans, and it never changes source code itself.

Ponytail forces the shortest working solution and favors YAGNI, stdlib, and native features over new abstractions. Use it when you want the simplest acceptable path or want to avoid over-engineering and extra dependencies.

Project Memory keeps a `.shob/memory/` folder of structured markdown files so project context survives across responses and sessions. Use it when you need to preserve state, resume work, or load only the relevant context for a task.

Figma Codegen turns a Figma selection into framework-aware code that matches the existing stack and styling. Use it when converting a Figma frame, screen, or component into React, Vue, Svelte, Next, or similar code.

Ask Matt recommends the smallest Matt Pocock skill or short skill sequence for a user's goal. Use it when you need help choosing the next skill, deciding between close options, or checking whether setup is required.

Domain Modeling clarifies concepts, language, boundaries, and architectural decisions when terminology is ambiguous or a design depends on shared understanding. It helps you compare terms, test edge cases, and record durable definitions or ADRs when needed.

Find Skills discovers installable agent skills from ctx's recommendation surface, the Skills.sh catalog, and the npx skills CLI. Use it to check whether a skill exists, find candidates, or safely evaluate updates before installing.