Back to the catalog

fahimtran

Inspired by Roo Code, has a suite of agents (and eventually skills) that aim to reduce context bloat and speed up task execution.

Open source Open in the app JSON README (API)

About

Inspired by Roo Code, has a suite of agents (and eventually skills) that aim to reduce context bloat and speed up task execution.

Details

Kind
Plugins
Topic
No topic detected
Publisher
fahimtran
Origin
marketplace
Category
ferramentas
Last push
2026-03-15T08:37:00Z
Repository state
ativo
Added
2026-08-30 01:48:58
Updated
2026-08-30 01:48:58
Origin id
fahimtran/claude-plugins/fahimtran

README

# claude-plugins

Personal Claude Code plugin — 14 specialized agents with an orchestrator-first workflow.

## Usage

### Load for a single session

```bash
claude --plugin-dir ./claude-plugins
```

### Install globally (user scope)

Requires a marketplace. See [Plugin marketplaces](https://code.claude.com/docs/en/plugin-marketplaces) to self-host one, then:

```bash
claude plugin install plugins@fahimtran
```

### Reload after changes

```
/reload-plugins
```

## How it works

When the plugin is active, `settings.json` sets `orchestrator` as the main agent. The orchestrator breaks down tasks and delegates to the appropriate specialized agent. Simple questions are answered directly without delegation.

## Agents

| Agent | Model | Role |
|---|---|---|
| **orchestrator** | inherited | Workflow coordinator — delegates to other agents |
| **architect** | inherited | System design and implementation planning |
| **ask** | inherited | Read-only Q&A and explanations |
| **code** | sonnet | Hands-on implementation (default workhorse) |
| **code-reviewer** | sonnet | Pre-commit code review |
| **db** | sonnet | Database schema, migrations, queries |
| **debug** | sonnet | Bug investigation and root cause analysis |
| **devops** | sonnet | Docker, CI/CD, infrastructure |
| **docs** | sonnet | Documentation writer |
| **pr-writer** | inherited | PR titles and descriptions |
| **refactor** | sonnet | Code restructuring without behavior changes |
| **researcher** | inherited | External research via web search |
| **security** | inherited | Security auditing (read-only) |
| **tester** | sonnet | Test writing |

"Inherited" means the agent uses whatever model is active in your Claude Code session.

### Design principles

- **Orchestrator-first**: `settings.json` activates the orchestrator as the main thread for all non-trivial tasks.
- **Parallel execution**: The orchestrator spawns independent subtasks (e.g. multiple file writes) as parallel agents rather than sequentially.
- **Model tiering**: Planning/research agents inherit the session model; implementation agents use sonnet for speed.
- **Permission modes**: Read-only agents (`ask`, `architect`, `code-reviewer`, `pr-writer`, `researcher`, `security`) use `permissionMode: plan` to prevent accidental writes.
- **Clear boundaries**: Each agent has explicit "What You Should NOT Do" guardrails.

## Structure

```
claude-plugins/
├── .claude-plugin/
│   └── plugin.json       # Plugin manifest
├── agents/               # 14 agent definitions
├── settings.json         # Activates orchestrator as main agent
└── README.md
```

## Requirements

- Claude Code >= 1.0.33

More