software-development-team-plugin

Software Development Agent Team Plugin

A Claude Code plugin that provides a 3-tier agent team for collaborative software development. The plugin includes 11 specialized agents and 11 matching skills that work together in a pipeline workflow: Plan → Research/Orchestrate → Parallel execution.

What This Plugin Provides

11 Specialized Agents

Tier 1 — Planning (read-only):

• planner - Strategic planner for task breakdown and roadmaps
• researcher - Research specialist for library docs, APIs, and specifications
• product-manager - Product manager for requirements and user stories

Tier 2 — Orchestration (read-only):

• orchestrator - Workflow orchestrator for delegation planning and parallel execution

Tier 3 — Execution:

• designer - UI/UX designer for component specifications
• software-engineer - Software engineer for feature implementation
• code-reviewer - Code reviewer for quality and security checks
• qa - QA engineer for testing and bug finding
• sre-devops - SRE/DevOps engineer for infrastructure and deployment
• security - Security specialist for vulnerability audits
• legal-compliance - Legal compliance specialist for regulatory review

11 Matching Skills

• /plan-task - Create structured implementation plans
• /orchestrate - Produce delegation plans for parallel execution
• /write-requirements - Write user stories and acceptance criteria
• /design-component - Design UI components with specifications
• /implement-feature - Implement features following project patterns
• /review-code - Review code for quality and security
• /write-tests - Write unit, integration, and E2E tests
• /deploy - Deploy applications and manage infrastructure
• /security-audit - Perform security vulnerability audits
• /compliance-check - Check license compatibility and regulatory compliance
• /research - Research libraries, APIs, and best practices

Pipeline Workflow

The standard workflow for complex tasks:

Step 1: /plan-task [description]
  → Forks to planner agent (read-only)
  → Returns structured task plan with dependencies
  → May include research tasks for unfamiliar libraries/APIs

Step 1.5 (if needed): /research [topic]
  → Forks to researcher agent (read-only)
  → Searches web, reads docs, gathers reliable information
  → Returns structured report with sources and recommendations

Step 2: /orchestrate [plan output]
  → Forks to orchestrator agent (read-only)
  → Returns delegation plan with parallel groups
  → Schedules research tasks in earliest group when present

Step 3: Execute delegation plan
  → Main session spawns Tier 3 agents in parallel per group
  → Groups execute in dependency order
  → Review agents run after implementation agents

Agent Architecture & Orchestration

The following diagram visualizes the 3-tier agent structure, agent-to-skill relationships, orchestration workflow, and parallel execution patterns:

flowchart TB
    subgraph User["User Request"]
        UserReq[Task Description]
    end

    subgraph Tier1["Tier 1: Planning (Read-Only)"]
        Planner[planner<br/>plan mode]
        Researcher[researcher<br/>plan mode]
        ProductMgr[product-manager<br/>plan mode]
    end

    subgraph Tier2["Tier 2: Orchestration (Read-Only)"]
        Orchestrator[orchestrator<br/>plan mode]
    end

    subgraph Tier3["Tier 3: Execution"]
        Designer[designer<br/>full access]
        SoftwareEng[software-engineer<br/>accept edits]
        CodeReviewer[code-reviewer<br/>plan mode]
        QA[qa<br/>accept edits]
        SREDevOps[sre-devops<br/>accept edits]
        Security[security<br/>plan mode]
        LegalCompliance[legal-compliance<br/>plan mode]
    end

    subgraph Skills["Skills"]
        SkillPlan[plan-task]
        SkillResearch[research]
        SkillOrchestrate[orchestrate]
        SkillRequirements[write-requirements]
        SkillDesign[design-component]
        SkillImplement[implement-feature]
        SkillReview[review-code]
        SkillTests[write-tests]
        SkillDeploy[deploy]
        SkillSecurity[security-audit]
        SkillCompliance[compliance-check]
    end

    subgraph Hooks["Lifecycle Hooks"]
        HooksJSON[hooks.json<br/>SubagentStart/SubagentStop]
    end

    subgraph Group1["Parallel Group 1: Implementation"]
        G1Researcher[researcher]
        G1SoftwareEng[software-engineer]
        G1Designer[designer]
    end

    subgraph Group2["Parallel Group 2: Review"]
        G2CodeReviewer[code-reviewer]
        G2QA[qa]
        G2Security[security]
    end

    subgraph Group3["Parallel Group 3: Fix"]
        G3SoftwareEng[software-engineer]
    end

    %% User flow
    UserReq -->|"/plan-task"| SkillPlan
    SkillPlan -.->|forks to| Planner
    Planner -->|plan output| Orchestrator

    UserReq -->|"/research"| SkillResearch
    SkillResearch -.->|forks to| Researcher
    Researcher -->|research report| Orchestrator