Old-rules review for Fable 5.2
Rules hard-coded for older models work against Fable 5.2: delegation, think-less, style-limit and ‘never’ lines quoted with a fix.
863 saves on X, as of
Read Anthropic's official Prompting Claude Fable 5.2 guide and review my project's CLAUDE.md, AGENTS.md, skills, subagent files and memory files. Identify standing instructions that tell you to delegate work to subagents, lines that tell you to think less or answer directly, hard-coded style limits, 'never' rules and old notes that have piled up, and approval rules that don't name the exact actions they cover.
For each issue, quote the original instruction, explain its potential impact, and suggest the smallest change needed. List any proposals involving safety, permission boundaries, or reduced verification separately for me to decide. Show me the recommendations first. Don't modify anything yet.Use it
Send the prompt and the guide’s link to Fable 5.2 in Claude Code, in your project.
You get Each issue with the instruction quoted, its potential impact and the smallest change, before edits.
Anthropic just published a prompting guide for Fable 5.2. You can ask Fable 5.2 to use it to review the old rules in your project.
In the guide’s own test, Fable 5.2 had a dozen notes from earlier sessions in context, and just one of them said “never retry.” That one line cut its retries after a busy tool from nearly every time to about 4 in 10.
The rules you hard-coded to stop older models from doing dumb things now work against the new one and burn tokens for nothing.
Send this prompt and the guide to Fable 5.2: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-2
Add Copybook to your agent, then ask it next time. Set it up
