Prompts, Skills, Memory and Model Parameters (＋ Menu)

This page describes the ＋ menu in the input field (More options): prepared prompts and skills (templates, one-click questions), the memory (permanent notes for the AI), the model parameters (temperature, answer length), including the page or issue as context, exporting and the temporary chat.

What this does

The ＋ button in the bar below the input field opens a small menu with everything that changes how a question is sent — without leaving the chat. Only one panel is open at a time; Esc or a click outside closes it.

Who can do this

All users. Memory, parameters and the saved settings are personal and visible to you alone.

Before you start

• The app is set up and the chat is open on any surface (see Getting Started).
• For prompts and skills: templates exist — provided by the administration or created by you as own skills.

Menu entries at a glance

Click ＋ (More options). The entries appear in this order:

Using prompts and skills

Purpose — use a prepared text instead of formulating it anew each time. Permission — all users. Before you start — templates exist: either provided by the administration (Providing Agents, Prompts and Skills for Everyone) or your own skills (Creating Your Own Agents and Skills).

Steps

1. Click ＋ and then 📚 Prompts & skills ….
2. Choose an entry. The list has up to two parts:
   • Skills — run with one click — the question is sent immediately; if an agent is assigned to the skill, that agent is used.
   • Templates — insert into the message box — the text lands in the input field so that you can still edit it.
3. If the entry contains placeholders (e.g. {{Project}}), a small form with the name of the entry and one field per placeholder appears. Fill in the fields.
4. Click ⚡ Run (skill) or Insert (template). Back returns to the list without running anything.

Expected result — for a template the finished text is in the input field; for a skill the question has been sent and the answer is being generated. Entries without placeholders are inserted or run directly in step 2.

Editing memory

Purpose — store details that should apply in every conversation: your role, your projects, your preferred style. Permission — all users; the notes are visible to you alone. Before you start — nothing.

Steps

1. Click ＋ and then 🧠 Edit memory.
2. Tick Memory active — sent with every request.
3. Write your notes into the text field (placeholder What should the assistant remember permanently? (e.g. role, projects, preferred style)), e.g. "I am a product manager on the ABC project and prefer short bullet points."
4. Click Save. The button is only active after a change.

Expected result — below the field the counter shows the length, e.g. 120/4000. The memory icon appears as a chip in the input area, and the menu entry reads 🧠 Edit memory (active).

Notes

• At most 4,000 characters; longer text is rejected with Memory is limited to 4000 characters.
• The notes go to the AI service with every request — in Jira and in Confluence alike, because memory belongs to your account, not to a product. Do not write anything confidential there that your organisation would not want to hand to its AI provider.
• If you clear the checkbox, the text is kept but no longer sent. To remove it completely, empty the field and save.

Model parameters

Purpose — adjust how freely and how long the AI answers. Permission — all users.

Steps

1. Click ＋ and then ⚙ Model parameters ….
2. Adjust the values:

1. Click Done — the values are saved and the panel closes. Reset puts both back to Default.

Expected result — as long as one of the values is set (not Default), a chip with a gear icon is shown in the input area (tooltip Model parameters adjusted — click to edit). The values are remembered for you in Jira and Confluence alike.

⚠️ Warning: Do not set Max response tokens too low. Some models reason internally first; if the budget is too small, it is used up before any text is produced and you receive the message The model used its token budget (…) on internal reasoning without answering — shorten the prompt or choose a non-reasoning model.

Including page context

Purpose — send the content of the current page or issue along automatically, e.g. for "summarize this". Permission — all users; the content is read with your own permissions. Before you start — the surface knows a page or an issue (issue panel, issue sidebar panel, issue action, macro, page assistant, issue navigator, or a fullscreen overlay opened from an issue or page — including the overlay opened from the Confluence chat bar), and the administration has not set AI access to Jira/Confluence to Off — chat without Jira/Confluence access. On the app page and the standalone full page the entry is not offered.

Steps

1. Click ＋.
2. Click Include page context. The checkbox icon shows whether the context is included; it is on by default.

Expected result — with every message the current content is sent along. The setting is remembered for you. In the issue navigator and for a question asked via the Confluence chat bar the context is always included, regardless of the setting.

What is sent

• For an issue: key, title, status, assignee, description and the last 5 comments (up to 800 characters each).
• For a page: title, space and the text content.
• In the issue navigator: up to 25 selected issues or your current search (JQL).
• Content longer than about 8,000 characters is cut off; the AI sees the marker [gekürzt, … Zeichen gesamt] (always in German, whatever the interface language) at the cut.
• The content is loaded fresh for every message — with your permissions. What you are not allowed to see, the AI does not see either.

Temporary chat

Purpose — ask something without the conversation being saved. Permission — all users.

Steps

1. Click ＋ and then Temporary chat (not saved).

Expected result — a fresh conversation starts and the chip Temporary ✕ appears (tooltip Temporary chat active — click to end). Clicking the chip or the menu entry again ends the temporary mode — and again starts a new chat.

Notes

• Switching deliberately starts a new chat: the previous conversation does not become temporary retroactively and stays saved.
• Temporary conversations do not appear under History and cannot be restored later. If you want to keep one, export it before you leave — via ＋ → ↓ Export conversation (Markdown) on the app page or ↓ Export in the header of dialogs and panels.
• Temporary chats still count towards the usage statistics and your hourly limit.

Common problems

Related topics

• Chatting with the AI
• Creating Your Own Agents and Skills
• Providing Agents, Prompts and Skills for Everyone
• Letting the AI Search, Read and Create in Jira and Confluence
• Managing Chat History: Pin, Fork, Archive, Delete
• Permissions, Roles and Data Visibility
• Troubleshooting
• FAQ