Local & open models

The top-rated launch-quality Local & open models skill for AI agents is react-native-ai-skills (grade B), followed by llama-cpp and gguf-quantization. Skillspace rates 161 Local & open models skills; 38 are launch-quality.

Running open-weight models: Ollama, llama.cpp, quantisation, local inference setups.

Launch-quality skills first, then by grade, then by our Skill Score, which adds judge reviews, test results, GitHub stars and adoption.

  1. react-native-ai-skills B
  2. llama-cpp B
  3. gguf-quantization B
  4. ollama-optimizer B
  5. sirvir B
  6. quantization-accuracy-tuning-orchestrator B
  7. turbofit B
  8. qwen-mtp-gguf C
  9. speculative-decoding C
  10. nemotron-nano3 C
  11. nemotron-super3 C
  12. llama.cpp C
  13. docker-model-runner C
  14. model-checker C
  15. afm C
  16. local-minicpm-pet-openvino C
  17. model-sovereignty C
  18. hal0-quantize C
  19. vllm-omni-setup C
  20. llama-cpp C
  21. transformers-to-mlx C
  22. discover-new-models C
  23. llamacpp-tuner C
  24. speech-vlm C
  25. llama-cpp-rpc-distributed C

Frequently asked questions

What is the best Local & open models skill for AI agents?

The top-rated launch-quality Local & open models skill for AI agents is react-native-ai-skills (grade B), followed by llama-cpp and gguf-quantization. Skillspace rates 161 Local & open models skills; 38 are launch-quality.

How does Skillspace rate Local & open models skills?

Skillspace grades every Local & open models skill from A to F with a published scoring engine that reads the skill's SKILL.md: architecture and maturity make the score, and a security scan can cap it. In crowded categories, a judge model also compares the strongest contenders side by side, and some skills are tested head-to-head on real tasks against other skills in the same category (including a deliberately weak one) and a no-skill baseline. Launch-quality skills are graded A to C with a clean safety verdict, and not excluded by a quality filter: the filters catch skills our model's safety screen flags for prompt injection, guardrail bypass, harmful content or credential exposure; test fixtures and attack benchmarks; skills written only for the repository they live in; templates and stubs; and near-duplicate variants and copies of other skills. Ratings were last updated on 24 September 2026; the full method is on the methodology page.

Are Local & open models skills safe to install?

Of the 161 Local & open models skills Skillspace rates, 130 passed its safety scan with no flag from its model, and 31 are flagged: 1 for review, 10 as risky, 19 as unsafe and 1 by its model for guardrail bypass. Each skill's page shows its safety verdict. A pass is not a guarantee: the scan reads only the skill's SKILL.md, not the scripts or other files it ships, and looks for patterns such as piping a remote script into a shell, reading credentials or hiding actions from the user. Skills are third-party code that Skillspace indexes but does not host or endorse, so read a skill's source on GitHub before you install it.

Skillspace indexes and rates skills; it does not host or endorse them. Run third-party skills at your own risk.