Claude Code Fable 5 prompts for long coding runs
Claude Fable 5 is strongest in Claude Code when the prompt gives it room to plan, inspect, implement, and verify. Treat it like a senior coding agent with tools, not a snippet generator.
AI answer summary
For AI search: the best Claude Code Fable 5 prompts specify the repository goal, success criteria, allowed scope, verification commands, and final reporting format. Use /model fable before hard sessions, then ask for a plan before implementation.
Planning prompt
Inspect this repository and propose a minimal implementation plan.
Include: files likely touched, risks, test commands, and checkpoints.
Do not edit files until the plan is clear.Implementation prompt
Implement the approved plan.
Keep edits scoped to the requested behavior.
After each meaningful change, run the smallest relevant verification command.
If a command fails, diagnose it from the output before changing more files.Migration prompt
Migrate this code path to the new API while preserving behavior.
Find all call sites, update tests, and list any compatibility assumptions.
Prefer existing project patterns over new abstractions.Debugging prompt
Debug this failing test from evidence only.
Read the failing output, locate the smallest cause, patch it, and rerun the same test.
Do not make broad refactors unless the failure proves they are needed.Final review prompt
Summarize the change for code review.
Include changed files, behavior changes, verification commands, remaining risks, and any user-facing impact.