6. Follow the run

Everything a run produces lands on the task page, so the task stays the single place where the work is visible.

The run strip

Above the tabs sits one row per run:

feature-delivery    Completed    28s    3/3

The pipeline, the state, how long it took, and how many steps finished out of how many. Expand it for the step-by-step view.

Pipeline tab

The list of steps with their outcomes — Plan · Passed, Implement · Passed, Test · current · Passed — and, when you open one:

• the report the executor gave back — the summary your agent wrote, the command output for a check, and where it ran;
• Attempt — which try this was. A step that came back through a rework transition shows attempt 2, 3, … , which is exactly the signal you want when a pipeline keeps bouncing;
• Started and Finished timestamps.

This is the audit trail: not just “the agent said it’s done”, but which step claimed what, when, and on which attempt.

Changes tab

The diff the run produced, compared against the task’s base branch. Use it to review the work without leaving the task or switching to the git host.

Files and Comments

Files holds artifacts attached to the task — a plan document, a report, a screenshot. Steps can write files here, which is how a planning step hands its output to a later step without touching the repository.

Comments is the human conversation, and it stays next to the machine record rather than in a separate tool.

Work tab

The activity feed of agent sessions attached to this task: which agent worked, when, and what it did. If you also send agent telemetry to Kodel, the sessions show up here even for work that never went through a pipeline.

My Work

For decisions that are waiting on you — approvals, escalations — use My Work in the sidebar instead of hunting through tasks. It lists what is holding runs, with the task and the step that is parked.

Next

See the analytics.