# Agent task board

Use Rowset as a small task ledger when agent work needs durable state across
runs, tools, and handoffs.

Task status is structured operational state, not a memory the agent may or may
not retrieve. See [AI agent memory vs structured
state](/blog/ai-agent-memory-vs-state) for the architecture boundary.

## Starter shape

Create an `agent_tasks` dataset indexed by `task_id`.

| task_id | title | owner | status | priority | blocker | completion_evidence |
| --- | --- | --- | --- | --- | --- | --- |
| TASK-104 | Draft onboarding copy | Scribe | doing | P2 |  | PR link required |
| TASK-118 | Decide API-key copy | Rasul | blocked | P1 | Needs product decision | Slack thread |
| TASK-121 | Verify export flow | Forge | todo | P2 |  | Test output |

## Agent jobs

- Create tasks with clear ownership and status.
- Move work only when dataset instructions allow it.
- Surface blockers across long-running agent sessions.
- Keep completion evidence attached to each closed task.

## Workflow rules

Define the allowed statuses up front: `todo`, `doing`, `blocked`, `review`, and
`done` are usually enough. Add instructions for who may move a task, what counts
as evidence, and when an agent should ask before taking action.

Because task agents often retry after tool or network failures, pair the board
with the [idempotent AI-agent update pattern](/blog/idempotent-ai-agent-updates).
Use `task_id` as the stable identity, write absolute status values, and verify
the row before replaying an uncertain update.

## Connect it

Use [MCP access](/docs/connect-mcp) for agent updates and the
[Dataset API](/docs/dataset-api) for scripts. Use public previews only for
read-only status sharing.