feat(opencode): add first-class OpenCode integration

[JasonOA888]

Implements #74

Summary

This PR adds comprehensive OpenCode integration following the same architecture as Claude/Cursor/Gemini integrations. All changes are production-ready with complete documentation and CI support.

---

🏗️ Architecture

Design Philosophy

• Consistency: Follows existing generation pattern (Claude/Cursor/Gemini)
• Automation: Auto-generated from existing metadata sources
• Validation: Built-in --check mode for CI pipelines
• Documentation: Complete integration guide

Data Flow

.claude-plugin/plugin.json  ──┐
gemini-extension.json      ──┼──> scripts/generate_opencode_config.py ──> opencode.json
skills/*/SKILL.md          ──┘

---

📦 Changes

New Files (336 lines)

File	Purpose
scripts/generate_opencode_config.py	Auto-generate opencode.json (205 lines)
opencode.json	Generated OpenCode configuration (131 lines)
docs/opencode.md	Integration documentation

Modified Files

File	Changes
scripts/publish.sh	Add opencode.json to generation pipeline
README.md	Add OpenCode setup section

---

📋 opencode.json Structure

{
  "name": "huggingface-skills",
  "description": "Agent Skills for AI/ML tasks...",
  "version": "1.0.0",
  "context": {
    "agents": "agents/AGENTS.md",
    "skills": "skills/"
  },
  "mcpServers": {
    "huggingface-skills": {
      "url": "https://huggingface.co/mcp?login"
    }
  },
  "skills": ["gradio", "hf-cli", ...],
  "skillDetails": [
    {"name": "gradio", "path": "skills/huggingface-gradio", "description": "..."},
    ...
  ],
  "metadata": {
    "author": {"name": "Hugging Face"},
    "homepage": "https://github.com/huggingface/skills",
    "repository": "https://github.com/huggingface/skills",
    "license": "Apache-2.0",
    "keywords": ["huggingface", "machine-learning", ...]
  }
}

---

✅ Key Features

1. Auto-Discovery

• Automatically discovers all skills from skills/*/SKILL.md
• Parses frontmatter for metadata (name, description)
• Falls back to directory name if no frontmatter

2. MCP Integration

• Reuses MCP configuration from gemini-extension.json
• Consistent authentication via Hugging Face Hub

3. Metadata Reuse

• Sources base metadata from .claude-plugin/plugin.json
• No duplication, single source of truth

4. CI Support

# Validate generated artifacts
./scripts/publish.sh --check

5. Complete Documentation

• Quick start guide
• Architecture diagram
• Contributing instructions
• Related integrations

---

🧪 Testing

Manual Testing

# Generate config
python3 scripts/generate_opencode_config.py

# Verify output
cat opencode.json

# Validate all artifacts
./scripts/publish.sh

CI Integration

# .github/workflows/ci.yml
- name: Check generated artifacts
  run: ./scripts/publish.sh --check

---

📊 Impact

Metric	Before	After
Agent integrations	3 (Claude, Cursor, Gemini)	4 (+OpenCode)
Generated artifacts	4	5
Documentation	3 agents	4 agents

---

🔗 Related

• Issue: [FEATURE]: add OpenCode skills configuration and generation pipeline #74
• Similar PRs: Cursor integration (Add OpenCode skills config and generation pipeline #75), Gemini integration

---

📝 Reviewer Notes

Design Decisions

1. Why skillDetails array?

   • Provides full metadata (path + description) for advanced use cases
   • Simple skills array for basic discovery
   • Both available in one config file

2. Why reuse .claude-plugin/plugin.json?

   • Single source of truth for metadata
   • Prevents drift between agent integrations
   • Follows established pattern

3. Why separate generate script?

   • Matches architecture of generate_cursor_plugin.py
   • Easier to maintain and test
   • Clear separation of concerns

Future Improvements

• Add OpenCode-specific skill triggers
• Support custom MCP server configurations
• Add usage analytics (optional)

---

This PR provides production-ready OpenCode integration with complete documentation, testing, and CI support.