OpenCode Config

Local configuration repo for OpenCode agents, skills, commands, and plugins. This is the source of truth for how your OpenCode instance routes requests and provides specialized behavior.

Structure

• agent/ - Custom agent definitions (Markdown with YAML frontmatter).
• skill/ - Reusable skills grouped by folder with a SKILL.md.
• command/ - CLI command definitions that route to agents (see command/gc.md, command/noslop.md).
• plugin/ - Local plugins (see plugin/notification.ts).
• opencode.json - MCP integration configuration (Notion, Jira, Gitlab toggles).
• AGENTS_IMPLEMENTATION_COMPLETE.md - Historical implementation note for the agent set.
• package.json, bun.lock - Dependency metadata for the config workspace.

Agents

• agent/orchestrator.md - Request router and delegation policy.
• agent/dev.md - TDD feature implementation and bug fixing.
• agent/code-review.md - Quality, security, and performance review.
• agent/code-pattern-analyst.md - Locate similar implementations and patterns.
• agent/docs-architect.md - Comprehensive documentation authoring.
• agent/oracle.md - Architecture guidance and best practices.
• agent/librarian.md - External research and documentation lookup.
• agent/ux.md - UI/UX design and frontend implementation.
• agent/commits.md - Conventional commit message generation.
• agent/commit-message-writer.md - Commit message drafting specialist.
• agent/fixup.md - Git fixup and autosquash guidance.
• agent/test-drop.md - Identify and remove redundant tests.
• agent/director_of_engineering.md - Senior engineering strategy and architecture.
• agent/project_manager.md - Delivery coordination and Jira-driven planning.
• agent/scrum_master.md - Agile ceremonies and flow improvement.

OpenCode also provides built-in agents like general and explore that are not defined in this repo.

Skills

• skill/conventional-commits/SKILL.md - Conventional commit message rules.
• skill/jira/SKILL.md - Jira workflow guidance and tooling.
• skill/gitlab/SKILL.md - GitLab CI/CD pipeline patterns.
• skill/fastapi-backend/SKILL.md - FastAPI backend patterns and JWT auth.
• skill/pytest-tests/SKILL.md - Pytest testing patterns and TDD advice.
• skill/golang-k8s-agent/SKILL.md - Go, gRPC, and Kubernetes patterns.
• skill/pyspark-databricks/SKILL.md - PySpark ETL on Databricks.
• skill/google-workspace/SKILL.md - Google Workspace API integration patterns.

Usage Conventions

• OpenCode loads agent/ and skill/ on startup; restart OpenCode after changes.
• Prefer explicit agent routing by name (@agent-name) for deterministic behavior.
• Keep agent frontmatter accurate: description, mode, tools, and permission control capabilities.
• Skills should be scoped, reusable, and focused on one job.

Adding an Agent

1. Create a new Markdown file in agent/ with YAML frontmatter (see agent/orchestrator.md).
2. Document the agent's role, triggers, and constraints in the body.
3. Register the agent in the orchestrator capability map so it can be routed (see agent/orchestrator.md).
4. Restart OpenCode to reload agents.

Adding a Skill

1. Create a new folder under skill/<skill-name>/.
2. Add SKILL.md with frontmatter (name, description) and usage instructions (see skill/conventional-commits/SKILL.md).
3. Keep the skill narrowly focused and runnable as-is.

Design Notes

• Clear separation of agent/, skill/, command/, and plugin/ keeps ownership boundaries explicit and reduces coordination overhead.
• Standardized frontmatter and folder layouts make behaviors discoverable, simplifying reviews and enabling consistent governance.
• Modular skills encourage reuse and safe composition, accelerating iteration without sacrificing reliability.
• A predictable structure lowers onboarding time and reduces architectural drift as the system scales.