Agents, Skills and Assistant

The Agents group in the Settings rail holds three pages: Agents, Skills and Assistant. They set how Neova talks to your agents, which instructions you can send them, and how the Assistant runs.

Agents

Open it from Settings › Agents.

Global Agents

A list of the agents in your global library. These agents are reusable across every project on this Mac. Each row shows the agent’s avatar, name and description, and an In this project badge when the open project uses it.

Button	What it does
Refresh	Reloads the list.
Reveal file	Shows the agent’s definition file in Finder.
Delete from library	Removes the global definition from this Mac after you confirm. Projects that use the agent keep their own copy, but it no longer receives updates and shows as a detached (orphaned) agent.

When the list is empty, it says to open an agent’s editor and choose Make global. To add a global agent to a project, or to edit it, use the Agent Library at the bottom of Agents › Team. See Team and Agent Editor.

Linking

Setting	What it does
Auto-link on name match	Links a terminal agent to an MCP agent automatically when their names match exactly.

History

Setting	What it does
History retention	How many days of agent activity history Neova keeps: 7 days, 14 days, 21 days or 30 days. The default is 7 days.

Event Messages

The messages Neova types into an agent’s terminal when something happens. The list on the left groups the events:

Group	Events
Chat	User messages, Agent messages, Mentions
Tasks	Task assigned, Task unassigned
Stages	Task entered stage
Agent Workspace	Question answered, Backlog reminder

Click an event to edit its message on the right. Under the text box, Variables: lists the variables this event can fill in. Click a variable to insert it at the cursor. If you type a {token} that is not a variable for this event, a warning shows: the token is sent to the agent exactly as written.

An event with unsaved changes shows a dot in the list. Click Save at the top of the section to save all changed messages. Reset puts one event’s message back to the default: click it, then click Confirm.

Which events an agent receives is set per agent in the Agent Editor.

Detection profiles

A detection profile tells Neova how to read a CLI’s terminal output, so it can show whether an agent is working, idle or needs your attention. The built-in Generic profile detects working and idle from the flow of output alone. Changes take effect immediately, without restarting the agent.

The Profiles list is on the left. Click New to create a profile, or use the arrow next to it and choose Add profile from catalogue. Starter patterns for common CLIs opens a set of ready-made Claude Code patterns with a Copy button for each.

Click a profile to edit it on the right:

Part	What it does
Name	Rename the profile. Saved when you leave the field.
Clone	Makes a copy under a new name.
Reset	Puts a profile that ships with Neova back to how it shipped. Only shown on those profiles.
Delete	Deletes the profile after you confirm. Not available on Generic, which every agent falls back to. A profile that agents use cannot be deleted; a dialog lists those agents.
Attention patterns	Text that means the agent is waiting for you, such as a permission prompt. Any match shows Needs Attention, even while output is still flowing.
Working patterns	Text that means the agent is producing output.
Byte-flow window (ms) and Byte-flow threshold	How much output, over what time, counts as working.
Idle patterns	Text that means the agent is back at its ready prompt. A match shows Idle right away.
Input ready patterns	Text that means the CLI is ready for the next message. Neova waits for it before it sends a message.
Ready-signal timeout (ms)	How long Neova waits for the input-ready signal before it sends anyway. From 500 to 30000. The default is 10000.

Each pattern list has Add pattern. A pattern has a name, an Enabled check box, a regular expression and a remove button. Patterns in one list are combined with OR: any one match is enough. Matching is case-sensitive. Enter the expression without surrounding slashes. Click Test against recent terminal output to try the expression on a running agent’s output.

Skills

Open it from Settings › Skills.

A skill is a reusable instruction for an agent. It can contain replacement fields such as {task.title}, which Neova fills in when you use the skill. The page has three tabs at the top: Global Skills, Project Skills and Quick Actions.

Global Skills and Project Skills

Global Skills are available in every project. Project Skills belong to the open project.

Both tabs show the same table:

Column	What it shows
Title	The skill’s name. Skills that ship with Neova show System default. A global skill from the catalogue shows Update available when there is a newer version.
Description	A short description.
Skill	The start of the skill’s text. Hover to see all of it.
Stages	The stages that use this skill as an activity.
Buttons	Duplicate skill and Delete skill. A system default shows a lock instead: it is read-only.

Above the table, Search skills… filters by name and description, and the stage menu (All Stages) shows only skills used in one stage.

Click a row to open the skill. A system default opens in View Skill, read-only.

On Global Skills, the arrow next to Create Skill has Add skill from catalogue. On Project Skills, Import Skills copies skills from another project folder: pick the folder, select the skills, and resolve any conflicts with skills you already have.

Skill fields

Create Skill, Edit Skill and Duplicate Skill show the same form. All three fields are required.

Field	What it holds
Title	The skill’s name, for example “Explain Task”.
Description	What the skill does, up to 120 characters. A counter shows how many you have used.
Skill Content	The instruction itself. The editor suggests replacement fields when you type `{`, and project files when you type `@`. Below it, a counter shows characters and lines. When you edit a skill, the file name shows too; click it to reveal the file in Finder.

For the full list of fields you can use in Skill Content, see Replacement fields.

How to create a skill

Open Global Skills or Project Skills.
Click Create Skill.
Fill in Title, Description and Skill Content.
Click Create Skill, or press ⌘S.

To change a skill, click its row, edit it and click Save Changes. To remove it, click Delete Template in the form or the bin in the table, then confirm. If you close the form with unsaved changes, Neova asks whether to discard them.

To make a skill appear on tasks in a stage, add it as an activity in Settings › Stages.

Quick Actions

Quick Actions are the skills you reach most often. They show:

in Search (⌘K), under Quick Actions, before you type,
as slash commands in the Chat input,
in the Templates › Quick Actions menu of the menu bar. Templates › Edit Quick Actions… opens this tab.

The tab has two lists:

List	What it does
Active Quick Actions	Your Quick Actions, in order. Drag to reorder. Click the minus to remove one. A skill that no longer exists shows as Missing skill; remove it with the minus.
Available Templates	All enabled global and project skills that are not Quick Actions yet. Click the plus to add one.

Each skill shows a Global or Project badge.

Assistant

Open it from Settings › Assistant.

The Neova Assistant runs a coding agent in its own terminal, so you can ask it about the page you are on. These settings apply to the open project. See Assistant for how to use it.

Neova Assistant

The header shows on or off. Changes in this section are saved when you click Save. Revert discards them.

Setting	What it does
Turn the Assistant on	Runs the Assistant. Saving a start command does not turn it on by itself.
System prompt	The Assistant’s behaviour, tone and rules. The text says whether it uses the packaged default or is customised for this project. Reset to default puts the default back.
Message template	What Neova types into the Assistant’s terminal each time you send it something. `{context}` is what you are looking at (view, target, cursor) and is left out when there is nothing to report. `{user_prompt}` is what you typed. Reset to default puts the default back.
Start command	The command you would type in a terminal to start your coding agent, for example `claude`, `claude --permission-mode auto`, `opencode` or `gemini`. It runs in this project with your terminal’s setup. The agent must support MCP.
Start automatically when this project opens	Starts the Assistant terminal in the background when the project loads. Saved per project. If Start command is empty, a warning shows.

Notices can show above the system prompt:

A newer standard prompt is available, with See what changed, when the catalogue has a newer default prompt.
A notice that this project’s prompt is based on an older version of the default. Click Reset to default to take the new one; your own edits are lost.

If the message template has no {user_prompt}, a warning shows and Neova uses the default template until you add the token back.

The Assistant can call any Neova MCP tool its agent has access to. Write in the system prompt when it should ask before it changes things.

Setup

A list of checks that must pass for the Assistant to work. The header shows ready or needs attention. Click Re-check to run them again. A failed check has a Fix button that opens the page where you fix it.

Check	Passes when
Start command	A start command is set.
Command found	The command’s program is found in your terminal’s PATH. Shows the path.
Neova MCP server	The MCP server is running. Shows the port.
Project MCP config	`.mcp.json` in the project points at this project.
Assistant registered	The running Assistant has connected to Neova.

A check that cannot be answered shows a question mark, not a failure.

Assistant terminal

The Assistant’s live terminal. You can type into it directly.

Part	What it does
Detection profile	The detection profile the Assistant uses. Saved when you change it.
Reset	Stops the Assistant terminal and starts it again with a fresh context.

When the Assistant is off, the area says Turn the Assistant on above to start its terminal.

How to turn on the Assistant

Check Turn the Assistant on.
Type a Start command, for example claude.
Optionally check Start automatically when this project opens.
Click Save. The header shows on, and the terminal shows under Assistant terminal.
Look at Setup. If a check fails, click Fix.

Assistant
Team and Agent Editor
Replacement fields
Settings › Stages to attach skills to stages
Settings › Server for the MCP server the Assistant uses
MCP tools such as list_skills, get_skill, assistant_reply and assistant_typing