Is this OpenCode-compatible?

[TomLucidor]

Wondering if this tool will be allowed for multiple coding agents that are more FOSS-oriented

[frmoretto]

Yes and no, I explain:
→ Yes (Soft Protection): Copy SKILL.md content into your project's AGENTS.md file. OpenCode will follow the safety rules as behavioral guidance. This works today but relies on LLM compliance (so it's not deterministic blocking).
→ No (Hard Protection): True pre-execution blocking requires a native OpenCode plugin using tool.execute.before hooks. This is on the roadmap... Looking for beta testers - if interested, comment here.

[TomLucidor]

Would like to beta-test and see where this goes

[frmoretto]

Great! A few quick questions so I can prioritize the right approach:

1. What OS are you running OpenCode on?
2. Any specific dangerous patterns you've encountered that you'd want caught?
3. Would you be comfortable testing the AGENTS.md approach first while I work on the native hooks?

I'll open a separate issue to track OpenCode support and tag you there.

[TomLucidor]

1. Let's all assume that Linux/OSX/WSL is the norm
2. The worst pattern is still rm work instead of using safer options like rip2 (and following Git best practices)
3. Will need to think of a few test scenarios for this

P.S. Another problem is premature halting, and pretending that work is "finished" by half-assing everything. Not sure if it is related to the other tool tho https://www.swyx.io/8020 https://xdg.me/80-20/ https://github.com/frmoretto/clarity-gate