This repo currently focuses on one skill: tlaplus-workbench. It helps convert natural-language designs into .tla/.cfg files, run TLC, and summarize counterexamples.
I would give it every example TLA+ file you can find, including PlusCal stuff (as reference). Also provide an up to date language manual and/or grammar if you have one.
I built a TLA+ skill for coding-agent workflows.
This repo currently focuses on one skill: tlaplus-workbench. It helps convert natural-language designs into .tla/.cfg files, run TLC, and summarize counterexamples.
Quick try: npx -y skills add younes-io/agent-skills --skill tlaplus-workbench
List skills from a local checkout: npx -y skills add . --list
TLA+ spec examples (generated from one-shot prompts): https://github.com/younes-io/tlaplus-workbench-examples
Feedback I'd value: 1. Is this useful for real protocol/state-machine modeling? 2. What features should I add next for this TLA+ skill?
I would give it every example TLA+ file you can find, including PlusCal stuff (as reference). Also provide an up to date language manual and/or grammar if you have one.
Do you think it helps with the systems reasoning? Are there many tla+ specs + real code bases for the llm to learn from?
Raft is a good example. And it seems to be in most models training data.