Copy-and-paste · verified July 27, 2026
Coding and debugging prompts for Claude Fable 5
Six scaffolds for refactoring, debugging and review, each anchored to something the model can check itself against.
What actually helps here
Make one command the definition of done. The documented failure mode of a long Fable 5 run is a confident, inaccurate progress report, and the fix is grounding: name the test command, the build, the linter — something with a real exit code — and require it to pass before the model reports finished.
The prompts
Fill in the bracketed slots, paste into any Claude app or the API. Written for Fable 5's goal-first style — they state the outcome and the constraints rather than scripting the steps.
Bug diagnosis
Here is an error and the relevant code: [paste]. List the three most likely root causes in order of probability, the fastest way to confirm each, and the minimal fix.
Code review
Review this function for correctness, readability, and edge cases: [paste code]. Point to specific lines, then show a revised version.
Refactor plan
Propose a refactor plan for [module description] that can ship in [3] small, independently testable steps. Note the risk of each step.
Test generation
Write [unit] tests for this function covering the happy path, boundary values, and one failure mode: [paste code]. Use [framework].
Explain code
Explain what this code does as if to a developer new to [language], then list two things that could break under load: [paste code].
API design check
Critique this API endpoint design for naming, status codes, and versioning: [describe endpoint]. Suggest a corrected spec.
Getting more out of these
Two habits matter more than the wording. Replace every bracketed slot before you send — Fable 5 follows instructions literally, and an unfilled [topic] becomes part of the task. And resist adding "think step by step" or "show your reasoning": on Fable 5 that phrasing can trip the reasoning_extraction classifier and hand your request to Opus 4.8 instead.
Need something these do not cover?
The generator builds a full system prompt around your own task, autonomy level and output style, and hands back the API call to go with it.