Make Claude Code work like a disciplined engineer: OODA, multi-party adversarial review, tiered model routing, fail-then-pass — token-efficient by design (route heavy work to smaller models, isolate sub-agent context). Distilled from Fable to reinforce the Opus harness.
Latest release v1.8.0 — Fable Harness v1.8.0 · 25 Oct 2026
7 files for Claude Code: red-team, simplifier, skeptic, adversarial-review and 3 more — 1,210 tokens loaded in every session.
CLAUDE.md A 1,210 tok .claude/skills/adversarial-review/SKILL.md A 142 tok .claude/skills/cognitive-rubrics/SKILL.md A 185 tok .claude/skills/model-dispatch-rules/SKILL.md A 149 tok .claude/agents/red-team.md A 71 tok .claude/agents/simplifier.md A 66 tok .claude/agents/skeptic.md A 66 tok These files are Miguok/fable-harness's own configuration — they tell Claude Code how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.