Slash commands & prompts

Type / in the AI overlay to open the command list: filter while typing, navigate with ↑/↓, complete with Tab/Enter.

Two markers qualify commands:

• 💬 interactive chat — opens a conversation session instead of answering in one shot.
• ♻️ replaces content — the command rewrites the note (or the selection).

Related commands are grouped in category folders (e.g. 📁 Spec kit); typing after / filters across all commands regardless of groups.

Built-in commands

Command	Kind	What it does
/brainstorm	💬	Interactive Q&A to shape an idea
/interview	💬	Interactive prep of a user-interview guide
/critique	💬	Devil's advocate: challenge your idea in a chat
/discuss	💬	Open discussion — structured summary at the end
/rewrite	💬 ♻️	Rework the whole note — replaces it
/edit	♻️	Rewrite only the selected text (one shot)
/summarize	—	TL;DR + key points of the current note
/eli5	—	Explain the note simply, from first principles
/specify	Spec kit	Turn the note into a structured spec
/clarify	💬 Spec kit	Interactive Q&A that removes ambiguity from the spec
/plan	Spec kit	Spec → technical plan
/tasks	Spec kit	Plan → implementation checklist with task IDs
/constitution	💬 Spec kit	Interactive interview to write the project constitution
/premortem	Spec kit	Imagine it shipped and failed — list why
/review	Spec kit	Review the spec: gaps, ambiguities, contradictions

Interactive sessions

Starting a 💬 command opens a transcript panel above the input — user/assistant bubbles, live status, one question per turn. Nothing is written to the note until you press the button.

• Generate draft writes the session's outcome at the end of the note; ✕ End session discards it.
• The session survives hiding the overlay with Esc; hover a bubble to copy it.
• The context toggles (current note, selected text, web search) and @ mentions work inside sessions too.

Replace commands — safe by design

/rewrite and /edit replace content under a strict contract:

• The replacement is applied in one editor transaction — a single undo restores the previous version.
• A reply that isn't a proper rewrite never touches the note: it lands in the chat instead, and the session stays open so you can retry.
• YAML frontmatter is preserved.

Custom prompts

Every command is a plain markdown file: <vault>/prompts/<command>.md — one file per command, versioned with the vault.

---
description: Translate the note into English
category: My prompts
interactive: false
replace: false
---
You are a professional translator…

• Any prompts/<name>.md whose name isn't a built-in becomes a brand-new /command, live immediately.
• An on-disk file for a built-in overrides it field by field — edit the prompt, its description, even make it interactive.
• Interactive commands keep their draft instruction in the same file under a ## Draft instruction heading.

Settings → Prompts

The Prompts tab in Settings manages the files:

• One row per command with its status: built-in (grey), on disk (orange), custom (blue), plus a version chip when a built-in default evolved since your file was generated.
• ↻ regenerates a single file from the current default; Reset all to defaults deletes the built-in override files (custom prompts are never touched).
• New custom prompt… scaffolds a ready-to-edit example file; Open folder reveals prompts/ in your file manager.