Simplify intro: remove negative scope disclaimer, state what it is

Also exclude chat UIs (ChatGPT, claude.ai, Gemini) from scope

Clarify scope: for building agents via APIs, not for prompting coding assistants

Update research sources: add Building Effective Agents and Advanced Tool Use, remove LangChain

Clarify two-step install in README

Fix plugin.json: author as object, remove skills field

Fix marketplace.json to pass validation

Add marketplace.json and README badges

Link Claude Code system prompts repo in README

Remove redundancies and outdated patterns from skill

Add cache-aware design, response format flexibility, think tool, and model-tier warnings

Initial release: prompt engineering skill for AI agent APIs