v0.1.0 — first public release

First public release of fable-skills — six Claude Code skills that harden Opus 4.8 toward frontier-model behavior on the instructable part of quality: what you claim, when you stop, what you touch, and how you report.

Skills

• fable-outcome-first — answer first; shape matches the question
• fable-finish-your-turn — finish reversible, in-scope work instead of asking permission
• fable-prove-it — claim only the rung you reached (written / runs / verified)
• fable-scope-discipline — build exactly what was asked; surface the rest
• fable-native-code — match the file's idiom; no defensive bloat or narrated restraint
• fable-context-thrift — spend context only on what changes the next action

How they were built

Distilled from Fable 5's behavioral contract, then TDD'd on the target model: pressure scenarios run on real Opus 4.8 subagents, failures captured verbatim, rules written against them, re-tested until they flipped. Full transcripts in docs/superpowers/testing/.

Honest scope

Proven: scenario-level behavior flips with transcripts. Not yet proven: a benchmarked daily-productivity gain (eval harness is on the roadmap). The baseline was Opus already running in a superpowers-loaded environment — these harden edges, they don't fix a broken model. Skills transfer judgment and discipline, not raw capability.

Install

Cross-platform: ./install.sh (macOS/Linux), .\install.ps1 (Windows), or copy the six skills/fable-* folders into ~/.claude/skills/. MIT licensed.