Personal Project using OpenCode CLI

This is a personal project with commands, skills and agents for the OpenCode CLI.

Core components

Agents are specialized AI workflows stored as markdown files in the agent/ directory. They define specific behaviors based on the task at hand. Whether agents operate autonomously depends on the tools and permissions available to them.

Sub-Agents are agents designed to be invoked by other agents (or by you) to handle focused subtasks. They provide specialized capabilities that primary agents can leverage.

Skills are packaged bundles of instructions, scripts, references, and assets stored in the skill/ directory. Skills extend OpenCode's capabilities by providing domain-specific knowledge, workflows, and tools. Think of them as "onboarding guides" that transform OpenCode into a specialist for particular tasks.

Core agents

OpenCode ships with built-in plan and build agents. They are defaults in the CLI and are not derived from HumanLayer.

HumanLayer-based commands and sub-agents

Commands and sub-agents adapted from: https://github.com/humanlayer/humanlayer/tree/main/.claude

Commands

• create-plan: Guides interactive implementation planning and produces structured plan documents
• research: Documents the codebase as-is by orchestrating parallel research agents
• implement_plan: Implements approved plans with phased verification and checklist updates

Sub-Agents

• codebase-locator: Locates files and components relevant to tasks
• codebase-analyzer: Analyzes implementation details and traces data flow
• codebase-pattern-finder: Finds similar implementations and usage examples
• thoughts-locator: Discovers relevant documents in the thoughts directory
• thoughts-analyzer: Extracts key insights from thought documents

Additional sub-agents

• code-simplifier: Simplifies and refines code for clarity and maintainability (sourced from .claude resources)

Skills

Skills are modular packages that provide specialized knowledge and workflows:

• skill-creator (from Anthropic): Complete workflow for creating, validating, and packaging new skills
• slack-gif-creator (from Anthropic): Knowledge and utilities for creating animated GIFs optimized for Slack
• steal-skill: Steal existing skills, commands, and agents from GitHub repositories
• agents-md-generator: Generates agents.md files for repositories with development instructions and best practices
• spec-writer: Helps write software specifications, requirements, and feature designs
• python-programming (testing): Python coding standards and workflow guidance with Google-style docstrings

Skills include their own documentation in SKILL.md files and may bundle scripts, references, and assets.

Commands

Custom commands for improved workflows:

• create-plan: Create implementation plans via the plan agent
• research: Document codebase as-is via the plan agent
• implement_plan: Implement plans via the build agent
• note-improvement: Capture agent/skill/command improvement suggestions and create GitHub issues

Getting Started

1. Install OpenCode CLI
2. Copy the agents, skills, and commands to your relevant opencode directories
3. Use OpenCode CLI to run agents and commands

For detailed instructions, see: https://opencode.ai/

License

This project is dual-licensed:

• Overall Project: MIT License - see the LICENSE file for details
• HumanLayer-inspired Agent System: Apache License 2.0 - see the LICENSE-APACHE file
• Skills: Licensed under MIT unless a specific LICENSE file exists in the skill's folder