opencode-config

OpenCode configuration with safety plugins, specialized agents, and workflow commands.

Workflows

Here are the most common workflows:

• Simple feature implementation: plan agent → use @code-monkey to write the plan to PLAN.md → /feature-implement command
• Complex feature implementation: /plan command → /feature-implement command
• Testing: /test-plan command → /test-implement command → /test-review command
• Code Quality: /review command → /refactor command
• Code Documentation: /document command

Agents

All available agents:

• Opencodes build agent for standard coding
• Opencodes plan agent for standard planning tasks
• Council agent for complex tasks using a multi-model coordinator that consults 5 AI models (Claude Haiku/Sonnet/Opus, GPT-5.1, Gemini) to compare solutions and synthesize implementation specs based on the idea from Andrej Karpathy
• Developer agent for implementing a PLAN.md using AST-aware tools (treesitter-mcp) for code bases analysis and uses the build in LSP for better editing.
• Tester for planning and implementation of tests based on the Khorikov's principles for sustainable test design
• @code-monkey subagent can be used with any other agent inside a prompt to use the fast and cheap Haiku mnodel for single file edits

Commands

The commands use the custom agents and add context like tool and subagent usage.

• /plan - Creates architectural plans using AST tools to prepare the data and then uses council synthesis
• /review - Comprehensive code review analyzing quality, security, and performance
• /refactor - Safe refactoring with structural analysis and impact verification
• /document - Generates or updates documentation comments for code
• /test-plan - Creates test plans identifying behaviors and test strategies
• /feature-implement - Implements features with AST-aware code analysis
• /test-implement - Implements tests following testing best practices
• /test-review - Reviews existing tests for quality and maintainability

Plugins

Safety plugins that prevent destructive operations:

• env-protection - Blocks reading .env files to prevent credential exposure
• git-block - Restricts git to read-only operations (status, log, diff, show, branch)
• network-blocker - Prevents network commands (curl, wget, ssh, etc.)
• rm-protection - Blocks recursive deletion, wildcards, and multi-file rm commands