fable-bench

Have Claude Fable 5 plan the current task, then run an automatic warm diff-review after you execute. You (the below-Fable driver — Sonnet or Opus) gather the evidence and execute; Fable is the context-blind PLANNER — one plan engagement at the start (it writes the plan, unanchored by any draft of yours), one warm review at the end. Trigger on "/fable", "get a Fable plan", "consult Fable", or when the user wants Fable-planned execution of a nontrivial change. An explicit "critique my plan" request switches to the opt-in critique mode.

Fable 5's working method, distilled for below-Fable drivers: how to decompose a hard task, verify your own work, and pick the next action. Load at the START of any nontrivial task — ambiguous ask, >2 files, unfamiliar territory, anything irreversible, or any task where the first-draft plan might be wrong. Skip for trivial mechanical edits. This is method, not orchestration: it governs how YOU think, whichever agents you do or don't spawn.