OpenCode Product Pipeline Skills

A complete 7-step multi-agent pipeline (8 skill files: 1 orchestrator + 7 agent skills) for OpenCode that transforms a raw customer idea into an implementation-ready project plan.

Overview

This skill collection implements the full product development lifecycle:

Step	Skill	Role	Output
1	interview-agent	Discovery	01-interview-brief.md
2	business-analyst	Business Analysis	02-business-analysis.md
3	product-manager	Requirements	03-product-requirements.md
4	ux-designer	UX Design	04-ux-design.md
5	solution-architect	Architecture	05-architecture.md
6	scrum-master	Planning	06-sprint-plan.md
7	developer-agent	Implementation	07-implementation-plan.md

The pipeline is driven by 8 skill files: the start-new orchestrator (the entry point — invoke it with /start-new) plus the 7 agent skills listed in the table above.

Installation

Option 1: Global Installation (recommended)

Copy all skill folders to your global OpenCode skills directory:

# Linux/macOS
cp -r start-new interview-agent business-analyst product-manager   ux-designer solution-architect scrum-master developer-agent   ~/.config/opencode/skills/

# Windows (PowerShell)
Copy-Item -Recurse -Path "start-new","interview-agent","business-analyst","product-manager","ux-designer","solution-architect","scrum-master","developer-agent" -Destination "$env:USERPROFILE\.config\opencode\skills\"

Option 2: Project Installation

Copy all skill folders to your project's .opencode/skills/ directory:

mkdir -p .opencode/skills
cp -r start-new interview-agent business-analyst product-manager   ux-designer solution-architect scrum-master developer-agent   .opencode/skills/

Option 3: Using npx skills (if published)

npx skills add <your-repo-url>

Usage

Start the Pipeline

Invoke the orchestrator directly with the slash command or its natural-language trigger:

/start-new

(or just type start-new / "start new product pipeline")

You can also simply tell OpenCode about your idea:

I want to build a medical OCR system for echocardiogram reports

The start-new skill will begin with the Interview Agent.

Or invoke a specific step:

Use the interview-agent skill to interview me about my product idea

Resume the Pipeline:

Continue the product pipeline

Go to a specific step:

Go to step 5 - design the architecture

State Tracking

The pipeline maintains state in pipeline-state.json in your project root:

{
  "current_step": 3,
  "completed_steps": [1, 2],
  "project_name": "Medical OCR System",
  "idea_summary": "...",
  "outputs": {
    "01-interview-brief.md": "completed",
    "02-business-analysis.md": "completed",
    "03-product-requirements.md": "in-progress"
  }
}

This allows the pipeline to resume across sessions.

Output Structure

All outputs are saved to a product-pipeline/ directory:

product-pipeline/
├── 01-interview-brief.md
├── 02-business-analysis.md
├── 03-product-requirements.md
├── 04-ux-design.md
├── 05-architecture.md
├── 06-sprint-plan.md
└── 07-implementation-plan.md

Skill Details

start-new (Orchestrator)

• Coordinates all 7 steps
• Manages state tracking
• Handles resume/rewind
• Ensures proper handoffs between steps

interview-agent

• Conducts structured interviews
• Uses open-ended → deep-dive → synthesis phases
• Produces a structured brief with problem statement, users, constraints

business-analyst

• Creates user personas (2-4)
• Performs competitor analysis
• SWOT analysis
• Market opportunity assessment

product-manager

• BRD and PRD/SRS documents
• Functional & non-functional requirements
• User stories with acceptance criteria
• Traceability matrix

ux-designer

• User flows (Mermaid diagrams)
• Text-based wireframes
• Screen inventory
• Accessibility considerations

solution-architect

• System architecture diagrams
• ERD (Entity Relationship Diagram)
• API design with schemas
• Security model
• Tech stack recommendations

scrum-master

• Epic breakdown
• Sprint planning (2-week sprints)
• Task breakdown with estimates
• Risk register
• Velocity projections

developer-agent

• Project structure
• Development setup guide
• Coding standards
• Implementation order
• Testing strategy
• Deployment guide
• Optional: Begin actual implementation

Customization

Each skill is a markdown file. You can:

• Edit the output templates to match your organization's standards
• Add additional sections to any deliverable
• Adjust the interview questions for your domain
• Modify the sprint duration or estimation approach

Compatibility

These skills follow the Agent Skills open standard — each SKILL.md carries the required name + description frontmatter, with name matching its directory. They are verified to load and trigger in OpenCode (launch the pipeline with the /start-new command). Other standard-compliant hosts — Claude Code, OpenAI Codex, Cursor, Windsurf — should load them as well; auto-trigger phrasing can vary by tool, so prefer invoking by skill name (or its slash command) where available.

License

MIT — see the LICENSE file for the full text.