How Skills Compare
Custom GPTs, Projects, saved prompts, and more
You might be wondering: “How are skills different from Custom GPTs? Or Projects? Or just… saving a good prompt?”
Fair question. Every AI tool has its own way of customizing behavior, and the names don’t always make the differences clear. Let’s break down the ones that cause the most confusion.
Pick what you’re curious about
Expand any comparison for a plain-language breakdown:
What it does
A custom chatbot with its own name, instructions, and personality. ChatGPT calls them Custom GPTs; Gemini calls them Gems. You launch them as separate "apps" from a sidebar — they're completely different conversations from your main one. Gems have the added advantage of deep Google Workspace integration, pulling from Gmail, Docs, Sheets, and Calendar natively.
How skills are different
Custom GPTs and Gems are separate assistants you switch to. A skill lives inside your existing assistant and activates automatically when the task matches. You don't have to remember which GPT or Gem to open — your AI just knows how to do the task because the skill taught it. Skills are also portable: the same SKILL.md file works across 40+ tools, while a Custom GPT only works in ChatGPT and a Gem only works in Gemini.
Analogy: A Custom GPT or Gem is like a specialist you visit at their office. A skill is like training your existing assistant to handle that job themselves, anywhere.
What it does
A block of text that applies to every conversation. Things like "Always respond in bullet points" or "I'm a marketing manager at a B2B SaaS company." These are global preferences — they shape how AI talks to you across all tasks. Code editors like Cursor and Windsurf have their own version of this (rules files), but the concept is the same.
How skills are different
Rules are always on. Skills activate only when relevant. This matters because cramming too many task-specific instructions into your rules makes every conversation slower and noisier. Rules should describe who you are and how you like to communicate. Skills should describe how to do a specific task. They layer together — your rules set the tone, and your skills handle the workflow.
Analogy: Rules are like telling a new hire "we use AP style and keep things casual." A skill is like handing them the step-by-step process for writing a campaign brief.
What it does
A reusable prompt you can quickly insert into a conversation. Like a text shortcut — you save a prompt once and recall it later instead of retyping.
How skills are different
A saved prompt is static text you paste in. A skill is a complete workflow with instructions, examples, quality checks, and sometimes reference files. A saved prompt says "write a report." A skill says "here's how we write reports, here's the template, here's a good example, here's what to check before you're done, and here's what format to output." The AI reads and follows the entire skill as a structured process — not just a starting prompt.
Analogy: A saved prompt is like handing someone a sticky note that says "make a report." A skill is like giving them your team's report playbook with examples and a checklist.
What it does
A workspace with uploaded files and a custom system prompt. Great for giving AI context about a specific area — your codebase, brand guidelines, product data, research documents.
How skills are different
Projects give AI context (what to know). Skills give AI capability (what to do). They're complementary, not competing. A project says "here are our brand guidelines and recent campaign data." A skill says "use those guidelines to write a social media post in our format." The project provides the raw material; the skill tells AI how to use it.
Analogy: A project is like giving someone a filing cabinet full of reference material. A skill is like giving them the playbook for what to do with it.
What it does
An automatic system that remembers details about you across conversations — your name, your role, your preferences, things you've told it before. Most major AI tools have some form of memory now, and it builds up passively over time so your AI feels more personalized the more you use it.
How skills are different
Memory remembers facts about you. Skills teach AI how to do a specific task. Memory might remember that you prefer bullet points and work in marketing. A skill knows how to write a campaign brief in your team's exact format with the right sections, tone, and quality checks. Memory is passive recall; skills are active capability.
Analogy: Memory is like a coworker who remembers you prefer bullet points and short paragraphs. A skill is like that coworker following a checklist to make sure the report gets done exactly the way you specified.
What it does
Connections that let your AI access external tools and data — Google Drive, Notion, Slack, Gmail, Jira, and more. Different tools use different names: Claude calls them "connectors," ChatGPT calls them "apps," and Gemini calls them "extensions." The underlying technology is often called MCP (Model Context Protocol), but you don't need to know that to use them.
How skills are different
Connectors give AI access to your data. Skills tell AI what to do with it. Connecting Google Drive lets your AI read your files. A skill tells it how to turn those files into a formatted weekly report. They solve different problems and work best together — connectors supply the raw material, skills supply the process.
Analogy: A connector is like giving someone access to your email, calendar, and project tools so they can send messages and edit documents. A skill is like giving them the process for what to send and how to organize it.
The short version
| Feature | Best for | Think of it as… | Portability |
|---|---|---|---|
| Skills | Repeatable workflows with standards | A playbook for a specific task | Cross-tool |
| Custom GPTs / Gems | Standalone chatbot experiences | A specialist you visit | One tool only |
| Rules / Instructions | Always-on preferences | Your style guide | Copy-paste |
| Saved Prompts | Quick reusable requests | A sticky note | Copy-paste |
| Projects | Context-heavy workspaces | A filing cabinet | One tool only |
| Memory | Passive personalization | Remembering your preferences | One tool only |
| Connectors | Accessing external tools and data | Access to your tools | One tool only |
These lines are blurring. AI tools are evolving fast, and features that are separate today may merge tomorrow. Custom GPTs may gain skill-like portability. Projects may absorb skill-like workflows. The important thing isn’t which feature you use — it’s that you’re building reusable processes instead of re-explaining yourself every time.
Takeaway: Skills aren’t competing with these features — they fill a gap none of them cover. If you need consistent output for a repeatable task that works across tools and across your team, build a skill.
Yes — and you probably should. Here’s how they combine:
Upload your skill to a project alongside your data files. The project gives AI your context, the skill tells it what to do with that context.
Use rules for your global preferences (tone, formatting defaults) and skills for specific task workflows. They layer on top of each other — rules set the foundation, skills handle the process.
Connect your AI to tools like Notion, Google Drive, or Slack, then use skills to process the data it pulls in. A “weekly report” skill becomes much more powerful when it can read your actual project data.
Memory remembers your preferences and context. Skills define your processes. Together, your AI knows both who you are and how you work — without you re-explaining either.
The best AI setups aren’t about picking one feature — they’re about layering the right features together. Skills handle the “how to do this task” layer that the other features don’t cover.
Need help bringing this to your team?
We work alongside your team to build AI-native workflows — from one-week sprints to full engineering acceleration. No handoffs, no slide decks.
Talk to us