agent-configuration

To configure OpenCode agents for custom workflows, you define prompts, models, tool access, permissions, and session behavior using mandatory YAML frontmatter with name and description fields. This enforces consistent agent behavior and governance across multi-agent orchestration sessions.

Setting up permissions and tool access for OpenCode subagents involves applying safety and permission configurations within the centralized agent definition. You constrain subagent capabilities during multi-agent orchestration by defining specific tool access boundaries and security rules in the configuration.

Yes, you can manage both primary agents and subagents within orchestrated OpenCode workflows. The configuration supports switching between primary agents and invoking subagents, allowing you to tailor prompts, models, and session behavior across the entire multi-agent orchestration.

YAML frontmatter enforces consistent behavior across OpenCode agent sessions by requiring a mandatory name and description for every agent definition. This standardized metadata ensures agents adhere to configured prompts, models, and governance rules throughout multi-agent orchestration.

You do not strictly need specific directories to configure OpenCode agents, but the setup supports optional directories like scripts/, references/, and assets/ for managing runtime resources. These directories help organize supplementary files required during multi-agent orchestration.

Security boundaries are important when configuring OpenCode agents for different teams because they enforce governance and constrain tool access. By defining explicit permissions and safety configurations, you prevent unauthorized actions and maintain consistent operational boundaries across multi-agent sessions.