> ## Documentation Index
> Fetch the complete documentation index at: https://docs.longoz.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Issues

> Create tasks, organize them in a list or board, and follow agent work through review.

An Issue keeps a task's requirements, assignment, activity, and results together. Create your first one with [Getting started](https://docs.longoz.ai/getting-started.md).

## Create a task

1. Choose **Create issue**.
2. Enter a title and select a [Project](https://docs.longoz.ai/projects.md).
3. Add a description and acceptance criteria. Explain what the agent should deliver and how you will judge it.
4. Assign a person or an eligible Project [Agent](https://docs.longoz.ai/agents.md), or leave it **Unassigned**.
5. Choose a status and create the Issue.

For agent-assigned Issues, choose **Todo** to request a Run or **Backlog** to park the task without starting one. An unresolved dependency can keep work waiting. The agent's [Computer](https://docs.longoz.ai/computers.md), provider, and workspace must be available for execution.

For example, in a Project containing a documentation repository, ask your configured CLI agent:

> Find broken internal links in the documentation. Fix links whose intended destination is clear. Report ambiguous links without guessing. List the files changed and the checks you ran.

Add acceptance criteria such as "No broken internal links remain in the pages you changed, and ambiguous destinations are listed in the summary." The repository and any link-checking tools you request must exist on the assigned Computer.

Use [Routines](https://docs.longoz.ai/routines.md) when you want a task created on a schedule.

## Choose a list or board

Open **Display** and select **List** or **Board**. Choose grouping by **Status**, **Assignee**, **Priority**, **Project**, or **No grouping**. Ordering options include priority, creation time, title, and update time. You can hide completed Issues and choose which properties appear.

Use **Filter** to narrow the view by Project, status, priority, label, assignee or Agent, work mode, and update dates. Search finds Issues within the current view. **Assigned**, **Created**, and **Subscribed** help you return to work relevant to you.

With the Board grouped by **Status**, drag a card to another status column to request that transition. The column's plus button opens Issue creation with that status preselected. Agent-assigned creation still requires Todo or Backlog.

Dragging changes status, not card order or assignment. Drops into non-status groups do not reassign an Issue. Some cards are locked by the current execution state or your permissions. Wait for a move to finish before moving another card. If the server rejects a transition, the Board restores the current server state.

Moving an Issue from **In review** to **Done** accepts its work. Read [Diffs and reviews](https://docs.longoz.ai/reviews.md) before doing this for code changes.

## Follow the status

Issues can be **Backlog**, **Todo**, **In progress**, **In review**, **Blocked**, **Done**, or **Cancelled**. Available transitions depend on assignment and the current Run. Status changes are execution actions for agent-owned work, so a running agent can lock property edits.

Open an Issue to read **Activity**, inspect its latest Run, or open **Session console**. The console becomes available after a Run exists. Select an earlier Run to inspect its output; input controls apply to the current Run only.

A Run's execution phase is separate from the Issue status. For example, an Issue can wait for capacity, a dependency, or a permission decision before the agent can continue. Resolve permission requests through their approval card, as described in [Inbox and approvals](https://docs.longoz.ai/inbox.md).

## Send context or continue work

**Add comment** saves discussion on the Issue. Use the composer action shown for its current state, such as **Send to agent**, **Send changes**, or **Send and retry**, when you want agent work. Check the delivery indicator: a saved comment is not proof that the agent received it.

Where available, **Add to queue** saves follow-up work for a later Run. The **Follow-up queue** shows whether each item is pending, running, completed, or paused. A failed or stopped Run can pause the queue. Use **Resume** when offered, or **Discard follow-up** to remove queued execution while keeping the original comment in Activity.

For an active Run, **Interrupt and send** stops the current work and saves your message. Continuation starts only after the Computer confirms the old process has stopped. Longoz uses a new provider session if the saved one is unavailable and records this in Activity. Existing queued messages stay paused while the interruption takes precedence.

**Stop** in the Session console stops the provider process. Once the Computer confirms it has stopped, the Issue becomes Blocked. A new comment or **Resume**, when available, continues the saved session. Closing an interactive provider terminal only detaches the terminal; the Run continues.

## Split work into sub-issues

Use **Create sub-issue** inside a parent Issue to track smaller pieces. Sub-issues inherit the parent's Project, assignee, and execution ownership. They do not start independent Runs or provider sessions. Use the parent Issue's execution controls to continue work.

For work that needs a separate agent or Run, create a top-level Issue. Add **Blocked by** relationships when it must wait for another Issue to finish. A cancelled blocker still blocks dependent work. **Related issues** records a relationship without expressing a blocking dependency.

Attach input files and download published outputs as described in [Artifacts and files](https://docs.longoz.ai/artifacts.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.