docs: Evaluate LangChain4j for SKILLS mechanism integration

[Copilot]

Evaluated whether to introduce LangChain4j's tool framework for Browser4's SKILLS mechanism.

Findings

Current State:

• SKILLS implementation in pulsar-agentic follows agentskills.io spec with progressive disclosure, resource-based structure (SKILL.md + scripts/references/assets), lifecycle hooks, and dependency resolution
• LangChain4j (v1.5.0) already integrated in pulsar-llm for LLM communication (chat models, caching, multi-provider support)

Evaluation:

Feature	Browser4 SKILLS	LangChain4j Tools
Progressive disclosure	✅	❌
Resource structure	✅	❌
Lifecycle hooks	✅	❌
Dependency management	✅	❌
Tool calling	✅	✅

Options Evaluated:

1. Replace SKILLS with LangChain4j tools → Loss of functionality
2. Use LangChain4j types for communication → Adds complexity, no benefit
3. Hybrid selective integration → Marginal utility, increased complexity

Recommendation

No integration needed. Current architecture is optimal:

• SKILLS mechanism exceeds LangChain4j tool framework capabilities
• LangChain4j appropriately scoped to LLM communication
• Aligns with project principle: "不引入外部编排框架（如 LangChain）— 先内部最小实现"

Documentation

Three evaluation documents added to docs-dev/copilot/:

• langchain4j-skills-integration-evaluation.md - Full technical analysis (487 lines)
• langchain4j-skills-integration-evaluation-zh.md - Executive summary in Chinese (328 lines)
• langchain4j-skills-integration-quick-ref.md - TL;DR reference

Optional enhancements suggested if improvements desired: JSON schema auto-generation, tool call format standardization, additional example skills, testing framework.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.