Opc_Kit
Professional AI Agent Skill Toolkit — From multi-persona decision debates to structured PRD output, a complete product workflow solution compatible with all AI coding agents
Why Opc_Kit?
Most AI tools give you a single perspective. Opc_Kit gives you a complete product decision loop.
- From Debate to Document — Use party-mode to simulate expert roundtables, then create-prd to turn decisions into structured, production-ready PRDs. Two skills, one seamless workflow.
- Eliminate Blind Spots — 17 professional personas covering engineering, product, and strategy challenge your assumptions before you write a single requirement.
- Professional-Grade Quality — Bidirectional traceability, first-principles validation, and 7-dimension scoring ensure every output meets senior PM standards.
- Zero Learning Curve — Natural language triggers with automatic intent detection. No commands to memorize, no configuration needed.
- Universal Compatibility — One skill set works across all AI coding agents with automatic tool adaptation. No vendor lock-in.
Skills Overview
| Skill | Purpose | Version | Quick Install | Guide |
|---|---|---|---|---|
| 📝 create-prd | PRD creation, update & validation | v2.5.1 | npx skills add sacrtap/Opc_Kit --skill create-prd | Usage Guide |
| 🎭 party-mode | Multi-persona product decision discussion | v1.0.0 | npx skills add sacrtap/Opc_Kit --skill party-mode | Usage Guide |
Quick Start
Installation
# Install all skills
npx skills add sacrtap/Opc_Kit
# Install specific skill
npx skills add sacrtap/Opc_Kit --skill create-prd
npx skills add sacrtap/Opc_Kit --skill party-mode
# List available skills
npx skills add sacrtap/Opc_Kit --list
Basic Usage
Use create-prd alone:
/create-prd Write a PRD for user authentication feature
Use party-mode alone:
/party-mode — Should we use microservices or monolith? We're a 5-person team at MVP stage.
Complete workflow: debate first, then document
# Step 1: Stress-test the idea from multiple perspectives
/party-mode — Should we build a real-time collaboration feature for our document editor?
Tech stack: React + Node.js. Scale: 5K concurrent users. Timeline: 8 weeks.
# Step 2: Turn the decision into a structured PRD
/create-prd Based on the party-mode discussion, write a PRD for real-time collaboration
Two skills, one seamless workflow: from multi-perspective debate to actionable documentation.
Skill Highlights
📝 create-prd — Professional PRD Writing Assistant
Transform product requirements into structured, production-ready documents with enterprise-grade quality assurance.
Key Features:
- 13-Chapter Standard Template — Fixed skeleton ensuring completeness, from problem description to risk analysis
- Bidirectional Traceability — US↔FR 1:1 mapping, every feature traces back to a user story
- Dual-Mode Workflow — Coaching mode (guided interaction, ~5-10 min) or Fast mode (direct generation, ~2-3 min)
- 7-Dimension Quality Scoring — Quantitative assessment with production-ready benchmarks (70+ = ready, 85+ = excellent)
- Exception-Covered Flowcharts — Mermaid diagrams with mandatory failure/timeout branches for all external calls
- Auto Language Detection — Chinese/English bilingual support with intelligent switching
Example Output:
User: /create-prd Help me write a PRD for user collection feature
Result:
✅ 13-chapter PRD with bidirectional traceability
✅ Mermaid flowcharts with exception paths
✅ Quality score: 82/100 (production-ready)
✅ Assumption index with 5 tagged inferences
📚 Read the full guide for detailed features, templates, and methodology.
🎭 party-mode — Multi-Persona Decision Discussions
Stop making critical decisions with a single perspective. Bring 17 product and engineering experts into real debates around your questions — not a pros/cons list, but a room of experts who argue, challenge assumptions, and push toward defensible conclusions.
Use it when the cost of being wrong is high:
| Scenario | What You Get |
|---|---|
| New product (0→1) | Stress-test market hypotheses, validate willingness-to-pay, identify real moats before committing resources |
| Key feature design | Catch UX gaps, security risks, scaling issues, and delivery blind spots before development starts |
| Architecture decisions | Debate irreversible choices with multiple experienced architects evaluating your specific constraints |
| Product roadmap | Evidence-based prioritization across competing priorities with multi-lens challenge |
| Business model | Unit economics validation, GTM strategy evaluation, long-term defensibility assessment |
Key Features:
- 17 Professional Personas — Engineering, product, and strategy experts with distinct expertise and communication styles
- Dynamic Role Selection — Auto-selects 4-6 most relevant experts based on your topic
- Three Discussion Tiers — Quick Take (3-5 rounds), Standard (8-12 rounds), Deep Dive (15-25 rounds) based on complexity
- Four Operating Modes — Subagent (independent thinking), Session (lightweight), Auto (hybrid), Agent-Team (persistent)
- Evidence-Driven — Roles cite industry data, benchmarks, and case studies, not just opinions
- Session Memory — Cross-session context retention, roles remember previous conclusions and alliances
Example:
/party-mode — We're building an AI habit tracking app. Target: professionals 25-40.
Hypothesis: $8/month for AI coaching. Budget: $200K, 4-person team, 6-month runway.
You'll hear:
- Cai challenging the $8/month assumption with real ARPU data for habit apps
- Ren asking if you've talked to 20 target users about their actual pain points
- Tao proposing a 2-week WhatsApp validation test before writing any code
- Wei warning that AI personalization is not a moat — every app will add GPT in 6 months
- Splinter questioning whether your target demo is even the right demographic
📚 Read the full guide for detailed scenarios, all personas, and advanced features.
Cross-Platform Compatibility
Skills are platform-agnostic by design. They use natural language instructions and generic tool descriptions that any AI coding agent can interpret and execute. No vendor lock-in, no configuration needed.
Verified Platforms
| Platform | Status | Notes |
|---|---|---|
| OpenCode | ✅ Full support | Native skill system, subagent support |
| Claude Code | ✅ Full support | Native skill system, subagent support |
| Cursor | ✅ Full support | Built-in tools, inline chat |
| Codex | ✅ Full support | CLI-based, full tool access |
| GitHub Copilot | ✅ Compatible | Workspace mode, chat interface |
| Windsurf (Codeium) | ✅ Compatible | Cascade flow, chat mode |
| Aider | ✅ Compatible | Chat-based interaction |
| Cline | ✅ Compatible | VS Code extension, full tool access |
| Continue | ✅ Compatible | Open-source, configurable |
| JetBrains AI | ✅ Compatible | IDE-integrated assistant |
| Amazon Q Developer | ✅ Compatible | CLI and IDE integration |
| Google Jules | ✅ Compatible | Agent-based workflow |
| Zed AI | ✅ Compatible | Built-in AI assistant |
| Void | ✅ Compatible | Open-source alternative |
| Trae | ✅ Compatible | IDE-integrated assistant |
How It Works
Skills follow a universal design pattern:
- Natural language instructions — Any LLM can understand the workflow
- Generic tool descriptions — "Read file", "Write file", "Search content" instead of tool-specific APIs
- Automatic fallback — When a feature isn't available (e.g., subagents), skills adapt gracefully
- No configuration — Just install and use, skills detect capabilities automatically
Adding Support for Your Platform
If your preferred AI tool isn't listed, skills will likely work out of the box. The key requirements:
- The agent can read and write files
- The agent can execute bash/shell commands
- The agent supports multi-turn conversations
That's it. No special integration needed.
Contributing
We welcome high-quality skill contributions!
Adding New Skills
- Fork this repository
- Create new skill folder (e.g.,
my-skill/) - Write SKILL.md following our structure guidelines
- Submit PR with usage examples
Skill Quality Standards
- ✅ Fixed template + mandatory validation mechanism
- ✅ Recommendation-driven interaction (not fill-in-the-blank Q&A)
- ✅ Bidirectional traceability assurance
- ✅ Professional perspective + industry best practices
- ✅ Cross-platform compatible (no Agent toolchain lock-in)
- ✅ Complete documentation + usage examples
License
MIT © sacrtap
Community
- GitHub Issues: Report issues or request features
- Discussions: Share use cases
- skills.sh: Browse and install skills
Opc_Kit — Empower AI Agents to become true product workflow experts, not simple Q&A machines.