MoonshotAI
Featured
Kimi-K3
Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.
fable-skills
fable-skills is a library of 35 composable instruction files that encode senior engineering habits—such as root‑cause debugging, security reflexes, and edge‑case sweeps—to lift LLM outputs toward Fable‑level discipline regardless of the underlying model.
Stars
93
7-day growth
No data
Forks
10
Open issues
0
License
MIT
Last updated
2026-07-10
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides auditable, benchmarked evidence (n=5 blind‑graded per cell) that adding these skill prompts to a lighter model like Sonnet 5 can match or exceed the bare performance of a stronger model like Opus 4.8, reducing variance and closing the model gap through explicit habits rather than model scale.
Who it is for
- Developers using Claude Code or similar LLM‑driven coding tools
- Engineering teams deploying LLM agents for code review, debugging, and planning
- AI researchers interested in prompt composition and model alignment
- Senior engineers evaluating structured vs. ad‑hoc LLM output quality
Use cases
- Improving code review thoroughness by activating security and correctness skills
- Guiding LLMs through structured root‑cause debugging with a reproducible checklist
- Enforcing edge‑case and numerical correctness when processing financial or concurrent logic
- Enabling a cheaper model to produce work equal to a more expensive one on specific tasks
Strengths
- Benchmarked with blind grading and full audit trail (62 raw outputs, answer keys, grader prompts)
- Skills are small, composable, and activated situationally—no monolithic prompts
- Proven lift across multiple models and task areas (debugging, concurrency, edge cases)
- Reduces output variance, making results more consistent and predictable
Considerations
- Benchmarks use n=5 per cell—directional but not statistically conclusive
- The strongest subtle insights (e.g., JSON null cache guard) still require the most capable model even with skills
- Grading is performed by an LLM applying an objective checklist, not by a human expert
README quick start
fable-skills
A library of 35 engineering-discipline skills — small, composable instruction files that encode how a careful senior engineer works, not what any particular framework does.
The premise
The most capable Claude models don't just know more — they work more carefully. Fable 5 reaches for a codebase's own conventions before writing, proves a bug's cause before fixing it, sweeps edge cases before declaring done, and reports what it actually verified rather than what it assumes. A strong-but-lighter model like Opus 4.8 can do all of that too — it just doesn't always reach for those habits unprompted.
These skills close that gap. Fable identified the places where a less deliberate run tends to fall short of its own ceiling — orientation, root-cause discipline, security reflexes, edge-case coverage, honest reporting — and wrote each habit down as an explicit, always-available checklist. Load them into a session and a model follows the discipline on purpose instead of leaving it to chance. The goal is simple: Fable-level output from whatever model is driving.
Each skill is one focused habit. They fire situationally — a debugging task pulls in root-cause-debugging, a diff touching user input pulls in security-reflexes, a schema change pulls in data-migration-safety — so the model carries the right discipline into the right moment without drowning in guidance.
Does it actually help? Benchmarks
Same method throughout: run an identical task through fresh agents with no guidance vs. with the relevant skill files, then score the outputs against a fixed answer key. Benchmark 1 uses Opus 4.8; Benchmark 2 runs both Opus 4.8 and Sonnet 5. These are directional signals, not statistical proofs.
The full assets are in benchmark/ — seeded tasks, answer keys, all 62 raw run outputs, the blind-grader prompts, and the per-item score tables — so every number below is auditable and re-gradable rather than take-my-word-for-it. (They were reconstructed verbatim from the original session transcript; each file carries a provenance header, and the recovery manifest in benchmark/README.md lists the known gaps.)
Benchmark 1 — Security code review (n=1, blind-graded)
The same task was run through two fresh agents and both reviews were scored blind by an independent third agent. Task: review a Python ord
Description
35 engineering-discipline skills authored by fable 5, with a blind-graded benchmark: sonnet 5 loaded with the skills beat opus 4.8 running bare. no deps, mit.
Related repositories
Similar projects matched by category, topics, and programming language.
xuchonglang
Featured
investing-for-beginners
A structured investing guide for Chinese beginners covering US stocks, options, and cryptocurrency, with focus on foundational concepts and risk awareness.
Krishnagangwal
Featured
CS-Fundamentals
A curated collection of Computer Science fundamentals (PDFs, notes, cheatsheets, interview question banks) for placement preparation, covering seven core subjects plus general resources.