Splitting epics into stories across four trackers

How Anvil Coder splits epics from YouTrack, Jira, GitHub and GitLab into stories, using configured triggers, separate coding approval and model usage.

What happens when an epic is triggered?

Anvil Coder can decompose a ticket marked as an epic into user stories. The project and tracker connection must be configured first. The epic enters the configured trigger state; the exact status or labels depend on the tracker and setup.

This does not start a coding run. Decomposition produces a proposal as tickets. It still uses models and may incur usage charges.

How do the four trackers differ?

GitHub uses df:type/epic to identify the ticket type; GitLab uses df::type::epic. The corresponding ready trigger is also required. YouTrack and Jira use configured ticket types, labels and statuses.

GitHub and GitLab use webhooks; YouTrack and Jira are polled. This is not a response-time guarantee: delivery, polling intervals and processing affect when work starts. See the integration overview for available connections.

How does an epic become a reviewable plan?

The system incorporates project foundations from the repository.
It clarifies the requirement and records assumptions.
It creates linked stories in the tracker without automatically triggering their implementation.
It reports the outcome or failure on the epic.

Review proposed stories for overlap, missing acceptance criteria and incorrect assumptions before approval. A structured proposal is still a proposal.

Why separate decomposition from implementation?

The team should be able to inspect the proposed scope before coding begins. Approving a story is therefore a separate decision. Budget for model usage during preparation as well as execution; “no coding run” does not mean “free.”

A versioned specification helps prepare larger projects. Small TaskSlices then concern implementation within a story, not the same planning step as epic decomposition.