Claude, ChatGPT, and Gemini each let you save reusable instructions so you don't rewrite your setup in every chat — but they store that context differently, and a portable skill file (a single markdown/text document) saves you the most friction on whichever one you switch to, or use all three at once.
Last updated: July 2026 · By the KissMySkills team
The problem: every platform "remembers" differently
If you've used more than one AI chat tool for real work, you've probably hit the same wall: you spend twenty minutes explaining your tone, your reporting format, your team's terminology, and your recurring tasks — and then the next chat starts from zero. Or worse, you build a great setup in one tool, the pricing or limits change, and you have to rebuild it somewhere else from memory.
Claude, ChatGPT, and Gemini all now offer some form of persistent context so you don't have to re-explain yourself every session. The mechanisms differ, though, and that difference is exactly where a portable skill file — one plain-text document you wrote once — starts paying for itself.
How each platform handles persistent context
| Platform | Where a skill lives | Notes |
|---|---|---|
| Claude | Project Instructions + Project Knowledge files, inside a Project | Instructions apply to every chat inside that Project; knowledge files sit alongside it as reference material. Large context window means a long, detailed skill file rarely needs trimming. |
| ChatGPT | Custom Instructions (global, account-level) or a Custom GPT (a packaged, shareable configuration) | Custom Instructions apply everywhere by default, which is convenient for one persona but awkward if you juggle several. A Custom GPT scopes a skill file to its own dedicated chat. |
| Gemini | Gems (saved, instruction-configured chat personas), with optional Google Workspace context | A Gem is essentially a saved system prompt you can reopen anytime. Workspace integration can pull in Docs, Sheets, or Gmail context automatically, on top of whatever instructions you set. |
Claude: Projects, Project Instructions, and a large context window
Claude organizes persistent context around Projects. Inside a Project, you set Project Instructions — a block of text that Claude treats as standing guidance for every conversation started in that Project — and you can attach Project Knowledge files that sit alongside it as reference material Claude can pull from.
This structure maps almost perfectly onto a skill file. You paste your skill file's content directly into Project Instructions (or upload it as a knowledge file, depending on length and how often you'll edit it), and every new chat in that Project starts already configured — no re-explaining tone, structure, or recurring tasks. Claude's context window tends to be generous relative to typical chat tools, which matters here: a well-written skill file with detailed examples, edge cases, and formatting rules doesn't need to be aggressively trimmed just to fit.
A single markdown file that turns Claude into a Chief of Staff persona — board packs, OKRs, decision memos, all-hands comms. Drop it straight into Project Instructions.
View Ryan — Chief of Staff →ChatGPT: Custom Instructions and Custom GPTs
ChatGPT offers two layers. Custom Instructions live at the account level — they're global, applying by default across new chats unless you turn them off for a specific conversation. That's great if you only ever want one persona active, but it gets messy fast if you switch between, say, a copywriting setup and a project-management setup during the same day: you end up editing the same global field back and forth.
Custom GPTs solve that by packaging instructions (plus optional files and configured behavior) into their own dedicated space, so you can keep several personas around and open whichever one you need without them colliding. Either way, the actual instructions you type are just text — which means a skill file written elsewhere pastes in cleanly, whether you're filling in the Custom Instructions box or the "Instructions" field when configuring a Custom GPT.
The practical friction with ChatGPT tends to show up when a team has several people each configuring their own Custom Instructions from memory, with small inconsistencies creeping in. A shared skill file — the same document everyone pastes in — keeps the setup identical across the team, and keeps it easy to update in one place and redistribute.
Gemini: Gems and Workspace context
Gemini's equivalent is a Gem — a saved chat configuration with its own instructions, which you can name, reopen, and reuse without retyping anything. Where Gemini adds a wrinkle is Google Workspace integration: a Gem (or a regular Gemini chat, depending on setup) can pull in context from connected Docs, Sheets, or Gmail, layering live document context on top of whatever instructions you've saved.
For a portable skill file, this means the instructions portion of your setup transfers over cleanly — you paste the same text into a Gem's instructions field — while the Workspace-context layer is additive, not something the skill file needs to account for. You get your standing persona from the file, plus whatever live document access you choose to connect on top.
Plain-text instructions for inbox triage, meeting prep, and agenda-building — paste into a Claude Project, a ChatGPT Custom GPT, or a Gemini Gem with no rewriting.
View Diane — Executive Assistant →Where the real savings show up
The token and time savings from a skill file come from two places, and they apply on all three platforms:
- You stop re-explaining yourself. Every chat that starts "cold" burns tokens (and your time) restating context a saved instruction set would have handled for free. A skill file loaded once into a Project, Custom GPT, or Gem removes that repeated overhead permanently.
- You stop rewriting the same setup per tool. This is the part people underestimate. If your skill file is written as plain, tool-agnostic instructions — no platform-specific syntax, no references to features unique to one app — it works as Claude Project Instructions, ChatGPT Custom Instructions, or Gemini Gem instructions without edits. Write it once, paste it three times.
That second point is where portability compounds. A skill file isn't locked into a proprietary format — it's markdown or plain text, the same kind of document you'd write in any text editor. That's what makes it copy-and-paste compatible with Claude, ChatGPT, and Gemini alike, instead of something you have to reformat or partially rebuild for each one.
Why portability matters beyond convenience
Two situations make this more than a nice-to-have:
Teams using more than one AI tool. Not everyone on a team standardizes on the same platform — some prefer Claude for long documents, some prefer ChatGPT for its ecosystem, some are already inside Google Workspace and default to Gemini. A shared skill file means the whole team runs the same configured persona regardless of which app any individual reaches for, instead of three slightly different, independently-drifted versions of "the same" setup.
Switching tools when pricing or limits change. AI platforms adjust pricing tiers, usage limits, and feature availability regularly. If your entire setup lives inside one platform's proprietary configuration — memory settings you can't export, or a Custom GPT you can't view the source of — switching tools means starting over. A plain-text skill file you own and control sidesteps that: it's just a document. You can paste it into whatever tool you're using next, today or in two years, without needing to reconstruct anything from memory.
In summary:
Claude Projects, ChatGPT Custom Instructions/GPTs, and Gemini Gems each give you a place to save persistent context, but only a plain-text skill file lets you use the exact same setup across all three without rewriting anything. Start with a ready-made skill like Ryan — Chief of Staff or Diane — Executive Assistant and paste it into whichever platform you're using today — and whichever one you switch to next.
Plug in how you (or your team) actually use AI chats and get a real monthly token-waste and dollar estimate for re-prompting vs. a persistent skill file. No signup, no email — just the math from this post applied to your numbers.
Try the calculator →Frequently asked questions
Can I really use the same skill file on Claude, ChatGPT, and Gemini without changing anything?
Yes, as long as the file is written as plain instructions — markdown or text describing a role, tone, and process — rather than using syntax specific to one platform. Paste the same content into Claude's Project Instructions, a ChatGPT Custom GPT's instructions field, or a Gemini Gem's instructions field.
Which platform saves the most tokens with a skill file loaded?
All three reduce repeated context-setting once a skill file is saved persistently, since you stop restating the same background in every chat. The exact savings depend on how often you start new conversations and how much context you'd otherwise retype — the mechanism (avoiding repetition) is the same across platforms, even if the underlying limits and context-window sizes differ.
Do I need a Claude Project, a Custom GPT, or a Gem specifically, or can I just paste instructions into a regular chat?
Pasting a skill file at the start of any single chat works, but it only lasts for that conversation. Projects, Custom GPTs, and Gems all exist specifically to make that configuration persistent, so you don't have to re-paste it every time — that's where the real time and token savings come from.
Related: Best Claude skills for executives & chiefs of staff