Why a story resists splitting
Almost every story that feels too big to estimate is actually several smaller stories wearing one title - the resistance usually comes from not having a pattern to split along, not from the work being genuinely indivisible. The patterns here (adapted from Mike Cohn's SPIDR framework, plus the common CRUD and happy-path variants) are the standard toolkit for finding that seam.
How to use the prompts
Pick two or three patterns that seem plausible for your story and work through their prompts - most stories split cleanly along at least one of workflow steps, business rule variations, or happy path vs. edge cases. If none of the patterns produce a clean split, that's often a sign the real blocker is unknown scope, not size - the Spike pattern covers that case directly.
When splitting isn't the answer
A story that's large because the approach is genuinely unknown - not because it bundles multiple deliverable pieces - doesn't need splitting, it needs a timeboxed spike to reduce the uncertainty first. Forcing a split onto unknown work often produces artificial pieces that don't actually reduce the risk.