NUX-Design GitHub avatar

claude-skills-fable-opus

NUX-Design

NUX Pack is a collection of six Claude Code skills, written using an early model and benchmarked to achieve 12 wins, 0 losses, and 2 ties in blind gradings.

Stars

23

7-day growth

No data

Forks

6

Open issues

0

License

No data

Last updated

2026-07-07

AI repository intelligence
FR-AI / ANALYSIS

Why it is worth attention

It presents reproducible, blind-benchmarked improvements for Claude Code, with a perfect win record and no losses across 14 evaluations, backed by full benchmark transparency.

Who it is for

  • Claude Code power users
  • Developers seeking to enhance LLM-based coding workflows
  • AI tooling researchers
  • Prompt engineers evaluating skill-based approaches

Use cases

  • Guiding Claude Code's planning and decision-making
  • Automating adversarial verification of outputs
  • Managing live state and truth consistency
  • Enforcing scope boundaries and applying ruthless editing

Strengths

  • Unbeaten blind-benchmark record (12–0–2)
  • Six distinct, task-specialized skills
  • Includes full benchmark methodology and rubrics
  • One-file installer for easy setup

Considerations

  • Not affiliated with or endorsed by Anthropic
  • Skills were written using a limited-access model (Claude Fable 5) that may no longer be available
  • Requires Claude Code environment and ~/.claude/skills directory setup

README quick start

NUX Pack v1.0.0

Six Claude Code skills written using Claude Fable 5 during its included-access window (July 1-7, 2026) and blind-benchmarked on Claude Opus 4.8. Aggregate: 12 wins, 0 losses, 2 ties across 14 blind gradings. Not affiliated with or endorsed by Anthropic.

What is inside

  • plan-gate/SKILL.md (2-0 in blind grading)
  • adversarial-verify/SKILL.md (1-0 (1 tie) + held-out win)
  • live-state-truth/SKILL.md (1-0 (1 tie) + held-out win)
  • scope-fence/SKILL.md (2-0 in blind grading)
  • ruthless-editor/SKILL.md (2-0 in blind grading)
  • memory-hygiene/SKILL.md (2-0 in blind grading)
  • install.sh one-file installer (inspect before running)
  • benchmarks.md the full method, every task, every rubric, every run including the losses

Install (pick one)

Manual: from inside this folder, create the skills directory (it may not exist yet) and copy the six skill folders into it:

mkdir -p ~/.claude/skills
cp -r plan-gate adversarial-verify live-state-truth scope-fence ruthless-editor memory-hygiene ~/.claude/skills/

Scripted: from this folder, run the installer (read it first):

bash install.sh

Claude Code picks up each skill automatically once its SKILL.md sits at ~/.claude/skills//SKILL.md.

Related repositories

Similar projects matched by category, topics, and programming language.

jamesob
Featured
jamesob GitHub avatar

local-llm

A comprehensive guide for building and configuring a high-end local machine to run state-of-the-art LLMs, with detailed hardware choices, BIOS tuning, and Docker-based model serving.

AI & Machine LearningLarge Language Models
1,660
bjan
Featured
bjan GitHub avatar

claudemux

Claudemux is a lightweight, dependency-minimal message bus that lets multiple Claude Code sessions communicate in real-time over tmux, enabling one session to ask another a question and receive the answer without manual intervention.

AI & Machine LearningAI Agents
9
AARomanov1985
AARomanov1985 GitHub avatar

Audio-Cassette-Simulation

A set of bash scripts using ffmpeg to simulate vintage cassette tape audio profiles with noise, wow/flutter, bandwidth limits, and EQ.

Shell
373