Slash commands, skills and prompts

The instructions you reuse, like a critique rubric, a summary format or a research method, shouldn't be retyped in every chat. Save them once as files, and / puts them one keystroke away on any canvas. The same menu also sets up the chat itself: persona, connectors and reasoning effort.

Open the menu

Type / in a chat node's composer or the side view, at the start of a line or after a space. Skills, Personas and Connectors appear as sections with a count (empty ones are hidden), followed by the reasoning options and your prompts.

Keep typing after / to filter. At the top level the search covers every section; inside a section it covers only that section. Skills and prompts match on name, description or file path, and reasoning, persona or connector lists everything of that kind. A space closes the menu, so search with one word.

What each item does

Personas, connectors and reasoning apply right away and leave no chip. They're the same settings as the + (Tools) menu in the composer.

Prompts

Prompts are Markdown files in ~/SlashspaceOS/.prompts, next to your canvases (~ is your home folder, C:\Users\<you> on Windows). Subfolders work too.

1. Create the .prompts folder if it isn't there. Finder hides names that start with a dot: press ⌘⇧. to show them.
2. Save a .md file in it, for example critique.md.
3. Type /crit in any chat. New files appear right away, no restart needed.

An optional header sets the prompt's menu name and subtitle:

---
name: Critique
description: Score a draft against my rubric
---

Critique the draft below. Rate clarity, argument and evidence from 1 to 5,
quote the weakest sentence in each area, and suggest one fix for each.

Without name, the menu shows the file name; without description, the file path. Other header fields are ignored, and the header is never sent.

To use it, insert /Critique, then paste your draft after the chip. Your message shows only the chip, but the model reads the full prompt. The file is read each time the chat sends, so edits apply from your next message.

Skills

A skill is a folder containing a file named SKILL.md that teaches the AI a workflow. The menu lists every skill in ~/.agents/skills, the shared folder the skills.sh installer uses, so skills you installed that way for other AI tools show up here too. Install more from Agent Skills in Settings.

Skills work like prompts: pick one to insert a chip, and its instructions go with your message. Only SKILL.md is sent, not other files in the folder such as scripts or examples. The menu name comes from the skill's name field, or its folder name.

Which modes use what

Reasoning effort works in every mode except Orchestrator. Outside Agent mode, the Connectors section shows "Available in Agent mode". Orchestrator mode runs in Slashspace's cloud on a fixed model: the menu still opens, but nothing in it changes what Orchestrator does.

If something goes wrong

• "Missing slash command reference": the file behind a chip was moved, renamed or deleted. Remove the chip and add it again. Editing the name field is safe; renaming the file or folder is not.
• A chat stops sending after you delete a file: an earlier message in that chat, or in a chat that feeds into it, still uses it. Put the file back.
• "Too large": one skill or prompt can hold up to 120,000 characters, and one message can expand to 400,000.
• Missing from the menu: type part of the name, since long lists get cut short. Symlinked files and folders are skipped.