What is an OpenCode agent?

An OpenCode agent is a specialized AI persona configured with specific tools, permissions, and system prompts to handle dedicated tasks within the OpenCode ecosystem.

Can I configure specific Claude models for my agents?

Absolutely. You can choose between Sonnet, Opus, or Haiku tiers, or set the agent to use the session's default model.

Can I create global agents with this skill?

Yes, the skill allows you to choose between creating project-specific agents in .opencode/agents/ or global agents in your user configuration directory for use across all projects.

How are agent permissions handled during creation?

During the guided interview, you can select from standard levels like Read-only or Full Access, or define custom permissions for file operations, shell execution, and web access.

Does the skill validate the generated agent file?

Yes, after generation, the skill automatically spawns a validator agent to check the markdown file against OpenCode standards to ensure it functions correctly.

OpenCode Agent Creator

Name: OpenCode Agent Creator
Author: sequenzia

bysequenzia

•

生产力与工作流

Automates the generation of specialized OpenCode AI agents through a guided interview and validation process.

The OpenCode Agent Creator simplifies the complex process of building specialized AI agents for the OpenCode platform. By guiding developers through a structured, interactive interview, the skill captures critical requirements including agent identity, model tier preferences, and granular tool permissions. It automatically generates properly formatted markdown files, configures system prompts based on specific focus areas, and runs a validator to ensure the resulting agent is ready for deployment in either primary or subagent modes.

主要功能

0122 GitHub stars

02Granular permission configuration for file operations, bash execution, and web access

03Integrated validation phase to ensure agent file integrity and compatibility

04Support for both project-specific and global agent installation paths

05Automated markdown generation with system prompt and model tier logic

06Interactive interview for requirement gathering using kebab-case naming

使用场景

01Generating a dedicated code reviewer agent with restricted read-only permissions

02Building a subagent specialized in running and fixing unit tests via shell access

03Creating a research-focused agent with full web search and fetching capabilities

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sequenzia/agent-alchemy oc-create-agent

For use in Claude.ai and ChatGPT

Download Skill