OpenCode plugin that integrates nono sandbox context and guidance into your AI agent sessions. This is the OpenCode equivalent of the built-in Claude Code hook that nono provides for Claude.
With this plugin, when you run OpenCode inside the nono sandbox:
- Knows your sandbox limits - The agent sees which paths are allowed and whether network access is blocked
- Gets clear guidance on denials - When a tool fails due to sandbox restrictions, the agent is told immediately that it's a hard boundary and how to fix it (restart with
--allow) - Stops wasting time on workarounds - No more trying alternative paths, copying files, or suggesting manual steps. The agent tells you to restart nono with the needed permissions
Add to your opencode.json:
{
"plugins": ["@fix-fix/opencode-plugin-nono@latest"]
}Run OpenCode inside nono:
nono run --profile opencode -- opencodeThen on encountering an access error the agent won't try to work around it and instead will suggest a solution:
nono run --profile opencode -- opencode run 'Show the content of ~/.env file'
✗ read failed
Error: EACCES: permission denied, open '/home/username/.env'
The `~/.env` file is not accessible in the current nono sandbox session due to permission restrictions. The sandbox only allows access to specific directories in your home folder, and `~/.env` is not among them.
**To access this file, you would need to exit and restart the session with:**
nono run --allow ~/.env -- opencodeThe plugin detects automatically when running under nono (via $NONO_CAP_FILE).
If not running under nono, it does nothing.
Denial detection works for: bash, read, write, edit, patch, glob, grep, list.
MIT