Skip to content

SuperCorks/subagents

Repository files navigation

Subagents

Specialized AI agent definitions for multi-agent workflows. Each .agent.md file defines an agent with specific responsibilities, capabilities, and instructions.

Installation

Use the @supercorks/skills-installer CLI:

Installer repository: supercorks/agent-skills-installer

npx @supercorks/skills-installer

Select the subagents you want to install. They will be placed in your chosen directory (e.g., .github/agents/ or .claude/agents/).

Available Agents

Workflow Agents

Agent Description
Orchestrator Coordinates multi-agent workflows, enforces quality gates, and ensures agents run in the correct sequence
Architect Designs implementation plans and technical architecture before any code is written
Developer Implements solutions defined by the Architect while adhering to project standards
Tester Ensures correctness and reliability by designing, reviewing, and executing tests
Code Quality Reviews implementation for maintainability, consistency, and standards compliance
Documentation Updates README, changelogs, and feature documentation
Git Manages branch state, validates repositories, and creates pull requests

Specialist Agents

Agent Description
Code Simplifier Refactoring specialist that reduces complexity while preserving functionality
Codebase Explorer Read-only analyst that maps architecture and traces execution paths
Security Auditor Red team agent that identifies vulnerabilities using OWASP patterns
UX Reviewer Design-focused QA for aesthetics, usability, and accessibility

Standard Workflow

The recommended workflow orchestrated by the Orchestrator agent:

Architect → Git (pre) → Developer → Tester → Code Quality → Docs → Git (post)

Workflow Stages

  1. Architect - Produces implementation plan, presents to user for approval
  2. Git (pre) - Validates branch state, ensures feature branches are ready
  3. Developer - Implements the approved plan, runs lint and tests
  4. Tester - Verifies test coverage, all tests must pass
  5. Code Quality - Reviews for maintainability, zero errors/warnings required
  6. Documentation - Updates relevant docs to reflect changes
  7. Git (post) - Creates PRs for affected repositories

Quality Gates

The workflow enforces gates between stages:

  • Architecture Gate: Plan must be approved by user before development
  • Git Gate (Pre): Must be on correct feature branch, up-to-date with develop
  • Implementation Gate: Code compiles, linting passes, tests run
  • Testing Gate: All tests must pass (zero tolerance policy)
  • Code Quality Gate: Zero errors and zero warnings
  • Documentation Gate: User-facing changes documented
  • Git Gate (Post): PRs created with semantic commit titles

Agent File Format

Each agent uses the .agent.md format with YAML frontmatter:

```chatagent
---
name: Agent Name
description: Brief description of the agent's purpose
tools: ['tool1', 'tool2', 'tool3']
---

Instructions

Agent-specific instructions here...


### Frontmatter Properties

| Property | Description |
|----------|-------------|
| `name` | Display name of the agent |
| `description` | Brief summary shown in agent selection |
| `tools` | Array of tool capabilities the agent can use |

## License

MIT

About

Collection of AI agent configuration files

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors