WenyuChiou/ai-research-skills

AI Research Skills

A skill catalog that won't let you spend 6 months on a research gap that's already closed, isn't a real contribution, or isn't feasible. Every candidate thesis topic goes through a 3-gate decision dossier first — and the rest of the pipeline (literature → design → build → run → write → submit) only flows from candidates that pass.

Languages: English | 繁中 · Pipeline · Examples · Runtime contract · Glossary · For Agent/Harness Builders

What this is. A catalog of 17 Claude Code skills built around one stubborn question most research-AI tools dodge: is this research gap actually worth doing? The pipeline opens with a structured decision dossier — three gates: open / contribution / feasibility — and downstream stages (research design, manuscript drafting, reviewer response) only fire on a candidate that clears all three. Built for graduate students, PhDs, postdocs, and research support staff. Five plugins, one marketplace install — and the same SKILL.md files load into Codex CLI, Gemini CLI, Cursor, Windsurf, Hermes, OpenClaw, and generic API clients too (see §7 Compatibility). For literature automation, the SKILL.md catalog and the executable research-hub runtime are separate layers; see docs/runtime-contract.md.

Optional harness extension. The 17-skill core is unchanged. Agent and harness builders can add the public agent-collab-harness policy/checkpoint layer; see the architecture and compatibility guide.

_{<a href="LICENSE"><img alt="License: MIT" src="https://img.shields.io/badge/license-MIT-blue"></a>}

📚 Part of the agentic AI learning roadmap — featured in §13–14 (research workflows).

---

Cont