How do I check a story before it goes into Jira, Azure DevOps, or Asana?
7 min read · Updated
Draft the story in your assistant, run the score-story command, and fix what the result names before the story exists in Jira, Azure DevOps, or Asana. Nothing downstream, no refinement slot, no plugin comment, no developer, ever sees the version that was not ready.
This is the workflow the Vindex MCP server was built for. Most stories already start in a chat window; the tracker gets a paste. Moving the check in front of the paste costs one sentence and turns the assistant from a drafting tool into a drafting tool with a reviewer in the room.
Why check before the tracker and not in it?
The tracker is where a story becomes other people’s work. The moment it exists in Jira it is in a backlog someone grooms, a filter someone reads, and a queue someone plans from. If the Vindex plugin is installed it will be scored there too, and that is valuable; but by then the story is already occupying attention, and the fix happens in a refinement session with six people watching instead of in a chat with one.
Checking first also changes what the tracker plugin sees. A story that arrives Healthy shows up green from its first minute, the team learns that green means something, and refinement time goes to the stories that need it.
How should I draft so Vindex reads it properly?
Keep the three parts of a story visibly separate in the chat: a title line, a description paragraph, and acceptance criteria as a list. The assistant passes them as three fields; if the criteria are buried in the description, Vindex still reads them, but the Testable reasoning is sharper when it can see the criteria on their own.
Draft in whatever format your team keeps. Vindex does not require “As a, I want, so that”; it reads for the outcome, the reason, and whether anyone could verify the result.
How do I run the check?
Either say it, “Score this story with Vindex,” or run the score-story command your client exposes. The command takes the story exactly as it stands in the conversation, whether you pasted it, the assistant drafted it, or you wrote it together, and sends it without changing a word. The result arrives finding first, then the overall percentage and its band, then the dimensions if you ask. Reading a result covers what each part means.
How do I change the story from the result?
One dimension at a time, starting with the finding. Story NW-105 from the Northwind sample backlog, “As a customer I want an ETA email when my delivery is delayed,” came back 62%, Needs review, through the live server. The finding: define the delay trigger and how the new delivery window is calculated. Small was the lowest dimension at 40, because the story bundled delay detection, the email, and an opt-out setting.
So the edit is two moves. Add a line to the description: a delivery is delayed when the ETA slips by more than 15 minutes, and the new window is the carrier’s revised ETA plus 30 minutes. Then cut the opt-out criterion into its own story. Ask the assistant to make those edits if you like; the judgment about what to change is yours, and Vindex’s reasoning tells you where to look. Vindex does not rewrite, and the command will not offer to.
Check again. The assistant keeps one story id for the story during the session, so re-checks are free, and you can go round as many times as the story needs.
When do I move it to the tracker?
When it is Healthy, or when it is Needs review for a reason you have decided to accept and can say out loud in refinement. Paste it into Jira, Azure DevOps, or Asana as you normally would. From then on, when you score the story again, give the assistant the tracker key; the assistant passes it as the story id, the plugin in the tracker uses the same identity, and the story is counted once wherever it is scored.
If your team runs coding agents, this is also the moment the story becomes safe to hand one. An agent with a ready check in its instructions will score it again before building; a story that left the chat Healthy passes that check without a second conversation.