Skip to content
Slicemap Docs

Your first story with the agent

View as Markdown to paste the page into your agent.

This page follows one story from the line you type to the moment the agent hands the story over. You need an agent connected with a Read & write key, and a map whose active release holds stories with acceptance criteria.

On its own, the agent takes stories only from the active release; a story you name by its number it takes from anywhere. The first release of a map becomes active on its own; after that, you activate the next release on the board or ask your agent to. If a map has no active release, the board says “The agent gets no stories until a release is activated”.

What it looks like

Section titled “What it looks like”
> take the next story from Slicemap


slicemap_next_story
  SM-12 "Class list with attendance", 3 criteria
slicemap_estimate_story
  40 min, 25 to 60, complexity 2, clarity 4
slicemap_update_story_status
  In Development
  ...writes the code, runs the tests...
slicemap_toggle_ac_done
  ✓ Coach marks who came to class
slicemap_toggle_ac_done
  ✓ Parent sees the kid's attendance history
slicemap_toggle_ac_done
  ✓ Coach sees who missed two classes in a row
slicemap_record_work
  35 min
slicemap_update_story_status
  Automated AI QA

Slicemap describes this order to the agent when it connects, and the agent follows it. Two rules are kept by the server whatever the agent does: the plan freezes when the story first goes into work, and a story can’t move to a closed stage while a criterion is unticked.

The board updates on its own as the agent works, so you can watch it in another window.

1. The agent takes the story

Section titled “1. The agent takes the story”

If your workspace has more than one project, name it: “take the next story from Slicemap in the Acme project”.

The agent calls slicemap_next_story. The map, not the agent, decides what is next: the first story of the active release that is not yet in a closed stage, going through the steps from left to right. The story comes with its acceptance criteria.

On the map: nothing changes yet. The agent now has the story’s number, title, description and criteria.

A story counts as open until it reaches a closed stage. A story the agent handed over and that waits in Automated AI QA comes back as the next story until you move it to Production, so review handed-over stories before you ask for the next one. When every story of the active release is in a closed stage, the agent stops and lists the backlog for you to choose from. It takes a story from outside the queue only when you name it, for example “take SM-142”.

2. It estimates the story

Section titled “2. It estimates the story”

Before the first line of code, the agent calls slicemap_estimate_story. It estimates in minutes of its own work, not a human developer’s days: the expected minutes with a range, how complex the work is (1 to 5), how clear the card is to it (1 to 5) and, below 5, one line on what the card is missing. Low clarity is a reason for the agent to ask you before it starts.

On the map: open the story. Its Agent estimate section shows the plan, the range, complexity and clarity, what is missing and the model that estimated. To see the estimate on cards, turn on Estimate under Display → Fields.

You can’t edit the plan yourself. If you disagree, ask the agent to re-estimate. More in Agent estimates in minutes.

3. It moves the story into work

Section titled “3. It moves the story into work”

The agent calls slicemap_update_story_status before it writes code. It picks the stage whose description says what is happening; with the default stages, that is In Development. From this moment the plan is frozen, so it can later be compared with the minutes the agent actually spent.

On the map: the card shows the new status, and on a Kanban tab it moves to that column. The card’s last change line reads like “Alex’s agent changed the status”, with the name of the person whose key the agent uses.

4. It ticks the criteria as it checks them

Section titled “4. It ticks the criteria as it checks them”

The agent calls slicemap_toggle_ac_done once per criterion, when it has checked that the criterion holds.

On the map: the counter on the card goes 1/3, 2/3, 3/3 and turns green when every criterion is ticked. In the story, the ticks appear under Acceptance criteria, and the last change line says the agent checked a criterion.

5. It records the minutes it worked

Section titled “5. It records the minutes it worked”

Before handing the story over, the agent calls slicemap_record_work with the minutes of its own work in this pass, without the time it waited for your answers. If the story comes back from acceptance, the next pass is added on top.

On the map: the Agent estimate section shows the actual minutes next to the plan, and the calendar time from the first move into work.

6. It hands the story over

Section titled “6. It hands the story over”

The agent calls slicemap_update_story_status again and moves the story on to acceptance or done. With the default stages, the stage after development is Automated AI QA, and the closed stage is Production.

If a criterion is still unticked, a move to a closed stage is refused:

A story is done only when every acceptance criterion is verified. Still unticked: "Coach sees who missed two classes in a row".

On the map: the card shows the new stage.

What next

Section titled “What next”