SOPs & process documentation
Turn how your team actually works into SOPs people can follow and agents can use.
Independent reviews. IonBin has no partnerships with the tools on this page and was not paid for these listings. Our evaluations are our own, based on public documentation and our research. Product names are trademarks of their owners and are used for identification only. How disclosure works.
Concept Stack 01 Stack 05
System prompt
Verified Oct 17, 2026 Review by Apr 15, 2027 4 sources
Short answer
A system prompt is the standing set of instructions a product gives a model before your message arrives: its role, rules, tone and the tools it may use. It shapes every answer. In your own setups, it is where your process belongs. In vendors' apps, it is usually hidden.
What it is
Every assistant runs with instructions you don’t type: who it is, what it must and mustn’t do, how to format answers and which tools exist. That’s the system prompt. When you build a custom assistant, you write it yourself. Anthropic’s docs note that even a single sentence setting a role focuses the model’s behavior and tone.
How it works
The system prompt is placed ahead of the conversation on every turn, so it governs every reply. Tool descriptions travel with it, which tells the model what actions are available.
Do vendors publish theirs? (checked 2026-10-17) Anthropic publishes the system prompts used in the Claude web and mobile apps, with dated updates per model. It notes these don’t apply to the Claude API. We didn’t find equivalent published prompts from other major vendors.
Benefits
- Consistency. Put the process, tone and checks in one place and every run follows them.
- Central updates. Change the instructions once and every user of that assistant gets the new version (see SOPs as custom assistants).
- Discoverability, if published. Simon Willison has called the system prompt and tool list “a detailed manual” for an assistant. He has complained that ChatGPT declines to reveal them, which makes features hard to discover.
Limits and risks
- Often hidden. Many assistants decline to reveal their system prompt, so users discover features by trial and error.
- Not a security boundary. Instructions guide behavior but don’t enforce permissions. Use access controls for that.
- Drift. Long, contradictory instructions degrade results. Keep them short and tested.
Best-fit stacks
Customer support
Answer faster from a help desk your agent can search, with help-center drafts a person approves.
Agent compatibility
| Tool in this entry | How agents connect |
|---|---|
| ChatGPT | Agent platform API/CLIBulk import |
| Claude | Agent platform API/CLIBulk import |
Levels come from each tool's recorded, sourced compatibility. How the levels work.
Related entries
How do I turn our SOPs into custom AI assistants? · QuestionHow do I write a prompt that gets usable work? · QuestionAgent harness · ConceptAgent skills (SKILL.md) · ConceptHow do I find out everything an AI tool can actually do? · QuestionHow do I get AI to ask me questions before it answers? · QuestionPrompt caching · Concept
Sources
- Giving Claude a role with a system prompt · Anthropicofficial accessed Oct 17, 2026
- Thread on discovering ChatGPT features and hidden system prompts · Simon Willison, 2026-08-02Practitioner view accessed Oct 17, 2026
- How to Get Ahead of 99% of People (with AI), custom GPTs section (from 16:29) · Dan MartellPractitioner view accessed Oct 17, 2026
- System prompts (release notes) · Anthropicofficial accessed Oct 17, 2026
Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.
Suggest an edit
IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “System prompt”? Email us with a link to a primary source.
Suggest an edit by email or write to [email protected]