Agent Skills by Wu Ben
Showing 23 vetted skills indexed across 1 GitHub repositories.
megumi-benidea-filter
Narrows broad research interests into crystallized, pipeline-ready paper ideas through constraint-driven filtering.
megumi-benidea-refine
Refines rough research ideas into venue-ready proposals via iterative external LLM review.
megumi-benidea-gen
Generates and ranks research ideas using external LLM brainstorming and multi-stage filtering.
megumi-benidea-screen
Screens and ranks research ideas via novelty checks, venue reviewer simulation, and strategic fit scoring.
megumi-benidea-pipeline
Orchestrates literature survey, idea generation, screening, and refinement into one autonomous research pipeline.
megumi-benlit-survey
Searches Zotero, Obsidian, local PDFs, and the web to build a literature landscape map with identified research gaps.
megumi-benopenbs-thesis
Guides QE reports, thesis proposals, final theses, and defense preparation following NUS guidelines.
megumi-benopenbs-job-search
Guides academic CS job search preparation, applications, interviews, and offer negotiation.
megumi-benopenbs-rebuttal
Drafts structured conference paper rebuttals from reviewer comments using categorized response strategies.
megumi-benopenbs-conference
Generates conference preparation plans with networking strategies, checklists, and email templates.
megumi-benopenbs-review
Reviews research paper drafts against Prof. Bingsheng He's quality checklists and five-figure rule.
megumi-benopenbs-submission
Validates academic papers against double-blind, COI, formatting, and submission integrity checklists.
megumi-benopenbs-collaboration
Guides research collaboration, industry meetings, and intern management using structured checklists and templates.
megumi-benopenbs-broader-impact
Generates Broader Impact sections for NeurIPS papers using a six-dimension framework.
megumi-benopenbs-idea
Guides research ideation and early PhD planning using Prof. Bingsheng He's survival framework.
megumi-benopenbs-camera-ready
Guides camera-ready paper preparation through a six-step academic review and formatting checklist.
megumi-benopenbs-presentation
Generates academic presentation outlines and reviews slides following Prof. Bingsheng He's guidelines.
megumi-benidea-create
Generates low-resource AI research ideas with top-conference potential using first-principles search.
megumi-benidea-eval
Evaluates research ideas with a five-dimension rubric and feasibility score.
megumi-benchatlog
Extract conversation messages from Claude Code or Codex session files into JSON.
megumi-bendrafting-research-plan-from-idea-folder
Generates a publication-grade research plan from a paper-idea discovery folder.
megumi-bencheck-idea
Analyzes research ideas and generates theory drafts, experiment plans, and critical reviews.
megumi-benwrite-paper
Writes NeurIPS-style LaTeX papers from existing idea, theory, and experiment materials.
Frequently Asked Questions About Wu Ben
FAQPage SchemaWhat tasks can I accomplish with Wu Ben's research skills?▼
You can run literature surveys with gap mapping, generate and rank 8-12 research ideas, screen ideas for novelty and venue fit, refine ideas into proposals, draft NeurIPS-style papers, write rebuttals and broader impact sections, and prepare camera-ready versions, presentations, and thesis documents.
Who are these skills designed for?▼
They target ML researchers, PhD students, and early-career academics—especially those following Prof. Bingsheng He's (NUS) methodologies—who need structured support for ideation, paper submission, conference preparation, academic job searches, collaboration management, and qualifying exam or thesis writing.
How do the idea-discovery skills work in practice?▼
Invoke idea-pipeline with a research direction and optional venue flag (ICML, VLDB, NeurIPS). It chains literature survey, idea generation via external LLM review, multi-dimensional screening, and deep refinement, producing ranked outputs like SCREENING_RANKED.md and a FINAL_PROPOSAL.md.
What dependencies or integrations do these skills require?▼
Core skills use Bash, Read, Write, Grep, Glob, WebSearch, and WebFetch. The full pipeline additionally integrates Zotero and Obsidian vault MCP servers plus Codex MCP for external LLM review. The openbs-* guidance skills are read-only and need no external services.
Do the skills support low-resource research constraints?▼
Yes. The idea-create skill performs first-principles search for low-resource AI research ideas with top-conference potential, explicitly accounting for strict time and compute budgets across foundation models, AI applications, and interdisciplinary directions.