Guide

Using the Catalog Effectively

This catalog currently focuses on Workflows — but Claude supports a spectrum of prompt types, from simple snippets to full agent configs. Understanding the differences helps you choose the right tool for each task.

Copy vs. Export — Which should I use?

Copy All Prompts

Paste directly into Claude

Assembles all the workflow's prompts into one block of text and puts it on your clipboard. Paste it into Claude's chat window and run the workflow start to finish.

Best for: One-off tasks, exploration, sessions where you won't need this workflow again for a while.

Export Markdown

Save to your project

Downloads the workflow as a .md file. Drop it into your project folder — Claude Code will pick it up automatically as context for every conversation in that project.

Best for: Ongoing projects where you want this workflow's approach to persist across every session without re-pasting.

The Prompt Type Spectrum

From a single reusable fragment to a fully autonomous agent — each type trades simplicity for power. Start simple; add complexity only when the task demands it.

Snippet

Skill

Workflow

System Prompt

Agent Config

← simpler · more complex →

Snippet

Simplest

A reusable fragment, not a full prompt.

Use when

You have a rule or instruction that applies across many different prompts — a constraint, a formatting preference, a reminder.

Example

"Always respond in markdown. Format code with TypeScript and add a comment on any non-obvious line."

Don't use when

The task needs multiple steps, or requires Claude to understand a broader context.

Skill

Simple

A self-contained prompt for one repeatable task.

Use when

You do the same task the same way, over and over. Write a commit message. Review a pull request. Summarize meeting notes.

Example

"/commit — Write a conventional commit message for these staged changes. One subject line, one short body paragraph. No bullet points."

Don't use when

The task varies significantly each time, or needs context that changes per-session.

Workflow

Moderate

A sequence of phases, each with its own prompt.

Use when

Complex tasks with distinct stages where each phase builds on the last. The output of one step is the input of the next.

Example

Research → Synthesize → Draft → Adversarial Review → Final Polish. Each phase has different goals, instructions, and success criteria.

Don't use when

The task is a single action or simple enough for one prompt — a workflow adds overhead without benefit.

System Prompt

Moderate

Persistent context and rules for an entire session.

Use when

You want Claude to behave a certain way throughout a whole conversation — specific expertise, tone, constraints, knowledge about your project.

Example

"You are a senior backend engineer on this codebase. Always prioritize security and readability. Never suggest adding a dependency without explaining why it's necessary."

Don't use when

The instructions only apply to a single message. System prompts set the stage; they don't direct individual steps.

Agent Config

Most complex

A full specification for an autonomous agent.

Use when

You want Claude to operate semi-independently across many steps, using tools (web search, code execution, file access) to complete a goal without step-by-step direction.

Example

A research agent: given a question, it searches the web, reads relevant pages, cross-references sources, and produces a structured report — all autonomously.

Don't use when

You want to stay in control of each step. Agent configs trade control for autonomy — only worth it when the task is long, repetitive, or tool-dependent.

Quick Reference

Type	Length	Phases	Autonomy	In this catalog
Snippet	A few lines	—	None	Coming soon
Skill	1 prompt	1	None	Coming soon
Workflow	Multi-prompt	2–10+	None	✓ Available now
System Prompt	1 block	—	None	Coming soon
Agent Config	Full spec	—	High	Planned