Prompt Templates
Build a library of reusable instructions and append them to prompts without retyping them every time.
Most people end up saying the same things to their agents over and over — use our test conventions, never touch the migration folder, write the commit message in this style. Prompt Templates let you write those once.

Building the library
Settings → AI → Prompt Templates is where templates live. Add, edit, hide, or remove them from one page. A template is just a name and a block of text.
Hiding a template keeps it around without offering it — useful for the one you only need during a migration.
Turning them on per terminal
Templates are opt-in per terminal. On any AI terminal you choose which enabled templates are appended when you send. A reviewer terminal and a builder terminal can carry completely different standing instructions.
Because it is per terminal, a template never leaks into an agent you did not mean it for.
Inserting one at the cursor
You do not have to append. In Agent Input, type @, open the Templates tab, and insert any enabled template directly where your cursor is. That is the right move when the instruction belongs in the middle of a longer prompt.
What to put in one
Templates work best for things that are true about your project rather than about this task:
- Coding conventions the agent keeps forgetting
- Files or folders that are off-limits
- The format you want plans, commit messages, or summaries in
- The test command and how to interpret its output
- A house style for comments and naming
Anything task-specific belongs in the prompt itself.
Templates vs skills vs project instructions
Three things overlap here, and they are worth telling apart:
- Prompt templates are 1DevTool's, appended to prompts you send from Agent Input. They work with every agent, including ones that have no instruction file of their own.
- Skills and slash commands belong to the agent and live in its own directory. See Skills & Slash Commands.
- Project instruction files (
CLAUDE.md,AGENTS.md, and friends) are read by the agent itself on startup and apply to every session in that repository.
Use a template when the instruction is yours and you want it applied selectively; use a project file when it should apply to everyone, always.
Related
- Agent Input — where templates are inserted and appended
- Skills & Slash Commands
- Project Configuration — shareable per-project setup