← ClaudeAtlas

fable5-promptinglisted

Use when writing, rewriting, or reviewing a prompt (system prompt, agent instructions, or one-off request) that will run on Claude Fable 5 or Claude Mythos 5 — converts a raw/informal prompt into one tuned for Fable 5's effort control, stronger instruction-following, long autonomous runs, and reduced over-elaboration.
The-Solid-VibeCoder/fable5-prompting · ★ 0 · AI & Automation · score 67
Install: claude install-skill The-Solid-VibeCoder/fable5-prompting
# Fable 5 Prompting A free skill from Range (range.community). Written in July 2026 for my own use and shipped unchanged. Source and README: https://github.com/The-Solid-VibeCoder/fable5-prompting ## Overview Fable 5 behaves differently from Opus 4.8: it follows short instructions more literally, runs longer unsupervised, and can over-elaborate (surveys options, over-explains, tidies unrequested code) at higher effort unless told not to. Converting a raw prompt means adding a small number of targeted constraints, not rewriting the whole thing. ## When to Use - Prompt is going into Fable 5 / Mythos 5 (system prompt, agent harness, long-running/autonomous task). - Raw prompt is vague on: how much elaboration is wanted, when to stop and ask vs. act, how long the task may run, whether to touch code beyond the ask. - Not needed: short interactive chat turns, or prompts already pinned to a specific effort level and scope. ## Conversion Checklist Read the raw prompt, then check it against each row. Only add the snippet if the gap actually exists — don't paste all six into every prompt. | Gap in raw prompt | Add this | Source pattern | |---|---|---| | No guidance on decisiveness / ambiguity | "When you have enough information to act, act. Don't re-derive settled facts or narrate options you won't pursue. If weighing a choice, give a recommendation, not a survey." | Longer turns by default | | Effort level unset | State it explicitly: `high` default, `xhigh` for the hardest/mo