🎯Skills79
An ARIS (Auto-Research-In-Sleep) skill that helps AI agents generate publication-quality academic figures, covering data visualization, plot formatting, and figure composition for research papers.
A research skill that systematically generates, validates, and ranks publishable research ideas by surveying recent arXiv literature, brainstorming with external LLMs, running lightweight pilot experiments, and scoring for novelty and feasibility.
A literature research skill from ARIS (Auto-claude-code-research-in-sleep), a zero-dependency workflow system that enables AI agents to autonomously conduct research, review papers, and iterate on scientific writing overnight.
A research automation skill that orchestrates a full idea discovery pipeline, chaining literature research, idea creation, novelty checking, and research review into a validated workflow.
An AI agent skill for automated academic paper research using Claude Code, providing autonomous arXiv paper discovery, scoring, weakness analysis, experiment execution, and narrative rewriting while you sleep.
Part of ARIS (Auto-Claude-Code-Research-In-Sleep), a lightweight zero-dependency system that lets Claude Code autonomously conduct research while you sleep — scoring papers, identifying weaknesses, running experiments, and rewriting narratives using plain Markdown skill files.
A skill within the ARIS framework that runs an automated iterative review loop — executing, scoring, identifying weaknesses, and rewriting research papers or code autonomously overnight, so users wake up to improved, higher-scoring outputs.
Part of ARIS (Auto-claude-code-research-in-sleep), a skill for autonomous ML research workflows that orchestrates cross-model collaboration—Claude Code drives research while an external LLM acts as critical reviewer, enabling paper scoring, weakness identification, experiment execution, and narrative rewriting autonomously.
A full paper writing pipeline that orchestrates outline planning, figure generation, LaTeX writing, PDF compilation, and automated review-improvement loops to go from a narrative research report to a polished, submission-ready PDF.
A research pipeline skill from ARIS (Auto-claude-code-research-in-sleep) that orchestrates autonomous multi-step research workflows — scoring papers, identifying weaknesses, running experiments, and rewriting narratives — using plain Markdown skill files compatible with Claude Code, Cursor, Trae, and more.
A skill within the ARIS (Auto-claude-code-Research-In-Sleep) framework focused on pixel art generation, allowing AI agents to create pixel art as part of a lightweight, zero-dependency research and creative workflow system built on plain Markdown files.
An autonomous research result analysis skill from ARIS (Auto-claude-code-research-in-sleep), a zero-dependency Markdown-based workflow system that lets AI agents score papers, identify weaknesses, and rewrite narratives overnight.
Deploys and runs ML experiments across local GPU, remote servers, Vast.ai, or Modal serverless environments, handling code sync, GPU pre-flight checks, and optional W&B integration.
Drafts LaTeX academic papers section by section from an outline, supporting major ML venues (ICLR, NeurIPS, ICML, CVPR, ACL, IEEE), with DBLP citation fetching, GPT-based section review, and optional style reference extraction.
Part of ARIS (Auto-Research-In-Sleep), a skill-based workflow for autonomous AI research that works as a Claude Code skill or standalone CLI. Performs novelty checking to evaluate whether research ideas are genuinely novel against existing literature.
Part of ARIS (Auto-Claude-Code-Research-In-Sleep), a skill-based workflow for autonomous AI research that monitors experiments, supports GPU deployment with code review, W&B logging, and targets major ML venues like ICLR and NeurIPS.
Turns a refined research proposal into a claim-driven experiment roadmap with a compact storyline of experiment blocks, ablation matrices, baseline selection, evaluation protocols, run ordering, and compute budgeting designed to defend paper claims.
Part of the ARIS (Auto-claude-code-research-in-sleep) autonomous research system, this skill provides Semantic Scholar paper search integration for literature surveys, enabling cross-model collaboration where Claude Code drives research while external LLMs provide critical review.
ARIS (Auto-claude-code-Research-In-Sleep) is a skill-based autonomous research workflow that runs Claude Code overnight to score research papers, identify weaknesses, run experiments, and rewrite narratives — implemented as zero-dependency plain Markdown files compatible with Claude Code, Cursor, Windsurf, and other agents.
Generates conference presentation slides (beamer LaTeX to PDF and editable PPTX) from a compiled research paper, with speaker notes and a full talk script. Supports multiple venues, talk formats, and aspect ratios.
Generates a structured, section-by-section paper outline from review conclusions and experiment results, with support for multiple academic venues (ICLR, NeurIPS, ICML, etc.) and optional style reference from existing papers.
An autonomous research workflow for Claude Code that iteratively scores, critiques, runs experiments, and rewrites academic papers while you sleep. Supports multiple LLM backends and works as both a skill-based workflow and a standalone CLI.
Generates publication-quality AI illustrations for academic papers using Gemini image generation. Employs a multi-stage workflow with Claude as supervisor to create architecture diagrams and method illustrations through iterative refinement, targeting CVPR/NeurIPS-style figure standards.
A proof-writing skill within the ARIS autonomous research framework that helps AI agents produce formal mathematical proofs and rigorous logical arguments during automated research sessions.
An ARIS (Auto-Research-In-Sleep) skill that implements a research-refine pipeline for unattended overnight research, using multi-executor and reviewer workflows with persistent memory, task tracking, and support for multiple LLM providers including Kimi, MiniMax, and GLM.
Part of the ARIS (Auto Research In Sleep) framework, enabling autonomous research workflows in Claude Code, Cursor, and Trae through skill-based automation.
Generates high-quality Mermaid diagram code from user requirements, supporting 23+ diagram types including flowcharts, sequence diagrams, class diagrams, ER diagrams, Gantt charts, and architecture diagrams, with syntax verification and file output.
Generates conference posters from compiled academic papers, producing A0/A1 PDF, editable PPTX, and SVG outputs. Supports venue-specific color schemes (NeurIPS, ICML, CVPR, etc.) with configurable layouts, column counts, and orientation for poster sessions.
Compiles LaTeX papers to submission-ready PDFs with automatic error diagnosis and fixing. Handles missing packages, undefined references, citation errors, and overfull boxes, supporting pdflatex, xelatex, and lualatex engines with up to 3 auto-fix attempts.
An ARIS (Auto-Research-In-Sleep) skill that runs automated review loops using MiniMax as the LLM reviewer, enabling unattended code quality improvement with multi-executor workflows and score-based iteration.
An ARIS skill for automated multi-round research paper review using LLMs, supporting API mode (fast) and browser mode (free) to iteratively improve paper quality toward submission-readiness.
A robotics-specific idea discovery pipeline that chains literature survey, idea generation, novelty checking, and critical review into an automated workflow grounded in embodiment constraints, simulator availability, and sim-to-real considerations.
An ARIS (Auto-Claude-Code-Research-in-Sleep) skill for Feishu/Lark notification integration with three modes (off/push/interactive), providing mobile notifications for research experiments, reviews, and checkpoints.
Bridges the gap between experiment planning and automated review by implementing experiment code from an EXPERIMENT_PLAN.md, deploying to GPU, and collecting initial results. Supports cross-model code review, sanity-first runs, and parallel experiment deployment for academic research workflows.
A Claude Code skill from the ARIS (Auto-Research-In-Sleep) framework that plans ablation studies for ML research papers, generating claim-driven experiment roadmaps with budgets and run order as part of autonomous research workflows.
Part of ARIS (Auto-Claude-Code-Research-In-Sleep), a lightweight zero-dependency system that lets Claude Code autonomously conduct research while you sleep — scoring papers, identifying weaknesses, running experiments, and rewriting narratives using plain Markdown skill files.
A single-agent literature review skill from the ARIS (Auto-claude-code-research-in-sleep) project that automates academic paper discovery, screening, and structured review using Claude Code's five-step research loop of plan, draft, adversarial review, iteration, and persistence.
Automates academic paper rebuttal writing by parsing reviewer feedback, enforcing coverage, and drafting venue-compliant responses for conferences like ICML.
A research workflow skill within the ARIS (Auto-claude-code-research-in-sleep) framework that transforms experimental results into formal claims. It is part of a four-layer evidence-and-claim assurance stack (experiment-audit, result-to-claim, paper-claim-audit, citation-audit) for validating and structuring academic research findings.
Checks the progress of autonomous research and training tasks running via ARIS (Auto-claude-code-research-in-sleep), a skill-based workflow for unattended AI research.
ARIS (Auto-claude-code-Research-In-Sleep) is a skill-based workflow and standalone CLI that enables autonomous overnight research with Claude Code. It features automated research loops, reviewer routing, and support for third-party Anthropic-compatible proxies.
A persistent research knowledge base skill inspired by Karpathy's LLM Wiki pattern. Accumulates papers, ideas, experiments, and claims with typed relationships across the entire research lifecycle, supporting ingestion, querying, syncing, and graph-based knowledge navigation.
A Claude Code skill that provides external patent examiner review of patent applications via multi-round Codex MCP review, offering critical feedback on claims and specifications using a patent examiner persona.
A skill for running GPU workloads on Modal, a serverless GPU cloud platform. Supports training, fine-tuning, inference, and batch processing with zero configuration — no SSH, no Docker, and automatic scale-to-zero billing.
Part of ARIS (Auto-claude-code-research-in-sleep), enabling Claude Code to autonomously conduct AI research overnight with automated paper scoring, weakness identification, experiment execution, and multi-executor orchestration across multiple LLM providers.
Orchestrates a full patent drafting pipeline from invention description to jurisdiction-formatted filing documents, chaining prior art search, novelty check, invention structuring, claims drafting, specification writing, review, and formatting for CN/US/EP jurisdictions.
Drafts patent claims with proper hierarchy, antecedent basis, and jurisdiction-specific formatting (CN two-part, US open, EP characterised-in-that), including cross-model examiner review and iterative revision.
Manages vast.ai GPU instance lifecycle — analyzes training tasks to determine GPU requirements, searches for best-value offers, and handles renting, setup, running, and destroying instances.
Assesses patent novelty and non-obviousness against prior art using element-by-element anticipation analysis, obviousness combination testing, and cross-model examiner verification for US, CN, and EP jurisdictions.
Part of ARIS (Auto-claude-code-research-in-sleep), a skill-based workflow that lets Claude Code conduct research autonomously while you sleep, featuring iterative scoring, auto-review, research wiki, self-evolution via meta-optimize, and multi-executor support.
Searches patent databases (Google Patents, Espacenet) and academic literature for prior art relevant to an invention, producing a structured report with overlap risk assessment and preliminary freedom-to-operate analysis.
Structures a raw invention idea into a formal invention disclosure using the Problem-Solution-Advantage patent framework, decomposing features into core inventive concept, supporting features, and optional features with cross-model validation.
A rigorous mathematical proof verification and fixing workflow that reads LaTeX proofs, identifies gaps via cross-model adversarial review, fixes each gap with full derivations, re-reviews until convergence, and generates a detailed audit report with proof-obligation accounting.
Part of the ARIS (Auto-claude-code-research-in-sleep) toolkit, this skill generates figure descriptions for academic papers and patent specifications, including reference numeral indexing and integration with the broader autonomous research workflow.
A patent drafting skill that writes detailed embodiment descriptions for patent specifications, covering method, system, and apparatus embodiments with proper reference numeral integration, claim support verification, and multi-language support (English/Chinese).
Audits experiment integrity using cross-model review (GPT-5.4) to detect common fraud patterns such as fake ground truth, score normalization issues, phantom results, and insufficient scope in research experiments.
Quick single-paper lookup using AlphaXiv LLM-optimized summaries with a tiered source fallback (overview, full markdown, LaTeX source) — explain or summarize arXiv papers from a URL or ID.
A Claude Code skill for compiling patent applications into jurisdiction-specific filing formats (CN/US/EP), supporting invention and utility model patents with Markdown and DOCX output for review and filing.
Verifies every bibliographic citation in an academic paper against three layers: existence at claimed sources, metadata correctness (authors, year, venue), and contextual appropriateness of the cited claims.
Performs AI-powered web search via Exa with built-in content extraction, supporting highlights, full text, and summaries. Useful for broad searches beyond academic databases, including blogs, docs, news, and company pages.
Zero-context verification skill that audits every number, comparison, and scope claim in academic papers against raw result files, using a fresh cross-model reviewer to prevent confirmation bias before submission.
A Claude Code skill providing paragraph-level structural blueprints for 10-12 page systems papers targeting top venues (OSDI, SOSP, ASPLOS, NSDI, EuroSys), with page allocation, paragraph templates, and writing patterns for each section.
Writes full patent specifications from claims and invention disclosures, generating all required sections (title, technical field, background, summary, drawings description, detailed description, abstract) with support for CN, US, and EP jurisdictions and automated claim-support verification.
Manages GPU compute jobs on the Qizhi platform using a kubectl-style CLI for job submission, monitoring, distributed training orchestration, and batch operations.
Generates publication-quality architecture, workflow, and pipeline diagrams as editable SVG from structured JSON specifications, ensuring deterministic and reproducible vector graphics output.
An SSH job queue skill for orchestrating large batches of ML experiments on remote GPU servers, with OOM-aware retry, stale screen cleanup, wave-transition race prevention, and multi-seed sweep support.
An automated research workflow (ARIS) that lets Claude Code conduct academic research while you sleep. Features multi-executor and reviewer support, persistent research wiki, meta-optimization for self-improvement, and a standalone CLI with support for local models via LM Studio/Ollama.
An automated research workflow (ARIS) that lets Claude Code conduct literature research, run experiments, and score papers while you sleep, with built-in research wiki, self-evolution, and multi-executor support.
A skill for searching academic papers via the OpenAlex API, providing access to open citation graphs, institutional affiliations, and funding data across 250M+ works from multiple sources.
Uses Gemini as an AI-powered broad literature discovery source that decomposes topics into sub-problems and searches across multiple angles and aliases, complementing arXiv, Semantic Scholar, and other research tools.
A two-thread adversarial review skill that stress-tests academic paper claims by constructing the strongest 200-word rejection memo, then defending point-by-point to surface unresolved critical issues before submission.
Automates academic research pipelines during idle time using Claude Code, managing literature review, paper reading, and resubmission workflows as background tasks.
An end-to-end conference talk pipeline that transforms a research paper into slide outlines, Beamer/PPTX slides, speaker notes, and full talk scripts with assurance checks for claims, citations, and anonymity.
Part of ARIS (Auto-claude-code-research-in-sleep), an autonomous ML research workflow system for Claude Code. This skill performs per-page visual review of presentation slides against a reference PDF, with fix patterns for font scaling, text-frame resizing, layout corrections, and anonymity checks.
ARIS (Auto-claude-code-research-in-sleep) is a skill-based research workflow that automates academic research using Claude Code and other AI agents through a five-step loop: plan, draft, adversarial review, iterate, and persist. It works across multiple AI coding tools including Cursor, Codex CLI, and Trae.
Part of ARIS (Auto-claude-code-research-in-sleep), a system that enables AI coding assistants to conduct autonomous research experiments while the user is away. Featured as a Hugging Face daily top paper, it supports skill-based workflows in Claude Code, Codex CLI, and other AI tools.
ARIS (Auto Research In Sleep) automates academic research tasks using Claude Code, enabling researchers to run literature reviews, data analysis, and paper writing overnight while they sleep.
Skill
Skill