Coding Standards
Reference standards for each language. Available as on-demand skills in .opencode/skills/ and as detailed reference files in .opencode/instructions/.
Agents load relevant skills on demand through the skill tool when work clearly matches a language/domain. Skills are not preloaded eagerly. Runtime behavior should prefer skills; instruction files remain a static reference layer.
Skill Activation Policy (OpenCode)
- Load skills on demand only when the task clearly matches a domain.
- Use one relevant skill by default; load a second only for explicit cross-domain dependencies.
- If stack/domain is unclear, clarify before loading.
Skill Permission Policy
Agents should define permission.skill with least-privilege rules:
- Start with
"*": "deny" - Add explicit
"<skill-name>": "allow"entries - Use wildcard patterns only when group-level access is required
Coverage
| Language | Highlights |
|---|---|
| .NET (C#) | Clean Architecture, async/await, nullable types |
| Python | Type hints, context managers, pytest |
| TypeScript | Strict mode, null safety |
| Flutter | Riverpod, freezed models, widget testing |
| Go | Modules, context, testing |
| Java | Spring Boot DI, validation, records |
| Node.js | Security middleware, validation, structured logs |
| React | Accessibility, performance, hooks |
| UX Responsive | Breakpoints, reflow, touch/pointer and keyboard accessibility |
| Ruby | MVC, ActiveRecord, RSpec |
| Rust | Ownership, Result/Option, clippy |
| CI/CD | Fail-fast gates, security, caching |
| SQL | Safe migrations, constraints, indexes |
| Blogger | Content style, SEO, research validation |
| Brutal Critic | Framework-based scoring, review process |
Full References
For detailed standards, open the corresponding file in .opencode/instructions/.
Execution Loop Guidance
- Use
.opencode/instructions/loop-execution.instructions.mdfor bounded iterative execution. - This protocol defines completion criteria, cycle limits, escalation rules, and verification gates.