.prompt.md files are reusable, version-controlled prompt templates for AI coding assistants — Markdown documents that package task-specific instructions (with optional frontmatter) so a common development task can be invoked on demand from a chat interface. They standardize repeatable work like code review, refactoring, and test generation.
.prompt.md
.prompt.md files are reusable prompt templates for AI coding assistants, stored as Markdown in a repository and invoked on demand from a chat interface. Each file packages a task-specific set of instructions — often with a small YAML frontmatter block for a description, mode, or tools — so a developer can run a common task the same way every time instead of re-typing a prompt. They turn ad-hoc prompting into a version-controlled, shareable asset that lives alongside the code.
- Markdown body plus frontmatter - Human-readable instructions with optional metadata that declares mode, model, or allowed tools.
- Repository-native - Checked into the project so prompts are versioned, reviewed, and shared like any other source file.
- Task-scoped - One file per repeatable job — code review, refactoring, scaffolding, or test generation.
- Invocable - Run directly from the assistant’s chat surface rather than pasted in each time.
In real API operations, .prompt.md files are where a team’s conventions get encoded for machines to apply consistently — a prompt that generates an OpenAPI operation to house style, drafts contract tests, or reviews a pull request against governance rules. As the agentic turn moves more of the development loop into AI assistants, these files become a governance surface in their own right: they capture how a team wants an agent to behave, keep that behavior repeatable across contributors, and sit close to the sibling patterns of Markdown-based agent instruction files.