The Agent Skills Directory

[SAFE]: The skill provides instructions and schemas for configuring agent permissions, modes, and tool access. It specifically guides users on how to restrict access to sensitive tools like bash and file editing using an allow/ask/deny model.
[SAFE]: No malicious patterns such as prompt injection, unauthorized data access, or remote code execution were identified. All paths (e.g., .opencode/agent/) and commands (e.g., opencode run) are standard for the intended agent management functionality.
[SAFE]: The examples provided demonstrate security best practices, such as creating read-only agents and restricting skill loading to specific prefixes.

opencode-agents