collaborative-deep-research/agent-papers-cli
4 skills1.4k installs192 starsGitHub
Install
npx skills add https://github.com/collaborative-deep-research/agent-papers-cliSkills in this repo
1Literature Reviewliterature-review is a Claude Code skill that runs systematic academic-style surveys on any technical topic via the paper and paper-search command-line tools. Before searching, the skill clarifies topic boundaries, year range defaulting to the last 5 years, target venues, and a paper count defaulting to 15-20 core papers. It executes multi-query searches with query variations to maximize coverage, then synthesizes findings into a structured review. Developers reach for literature-review when they need a survey, systematic overview, or evidence-backed summary of a research area before prototyping or writing technical documentation.692installs2Deep ResearchDeep-research is an agent skill that walks a solo builder through systematic topic investigation using the paper and paper-search CLI tools. It starts with wide web and Semantics Scholar queries, then refines with year, citation, and domain filters (including PubMed when biomedical). For the strongest hits, the agent outlines, skims, and reads sections of papers, browses key URLs, and walks citations and references to build context. It fits indie builders who need evidence-backed answers before scoping a product, picking a stack, or writing a spec—not a quick single-query chat answer. The workflow is procedural and repeatable, so you can hand the same ritual to Claude Code, Cursor, or Codex whenever a question deserves primary sources and traceable citations rather than model recall alone.296installs3Fact CheckFact-check is an agent skill from the agent-papers-cli stack that turns a single bold statement into an evidence-backed assessment solo builders can trust before they write landing copy, pitch decks, or technical posts. When a user asks to verify, fact-check, or confirm something, the skill first decomposes the claim into verifiable sub-claims, then runs structured paper-search queries across the open web and Semantic Scholar, drills into arXiv sections or browsed URLs for primary material, assigns one of four verdicts per sub-claim, and finishes with an overall assessment plus cited URLs. It fits indie researchers who cannot afford a human fact desk but still need to avoid publishing hallucinated statistics or misquoted papers. The allowed-tools surface (Bash, Read, Glob, Grep, Write) implies network-backed CLI search rather than guesswork from model weights. Use it in early Idea research, during Validate when testing positioning claims, or in Launch/Grow when auditing public-facing statements. It is not a legal compliance review or real-time news wire—latency and coverage depend on what paper-search returns.217installs4Research CoordinatorResearch Coordinator is a meta-orchestration skill for solo builders and agent users who receive open-ended research questions too large for one prompt. It sits above specialized workflows—deep investigation, systematic literature review, and targeted fact checking—and chooses the right path after analyzing intent. When the ask is compound, it decomposes work and dispatches multiple general-purpose subagents, each seeded with the appropriate skill file content and tooling access for academic paper and search CLIs. The coordinator synthesizes outcomes rather than replacing domain depth in child skills. It matches indie founders validating markets, engineers surveying prior art, or operators verifying claims before shipping copy. Ambiguity triggers a clarify-first gate so you do not waste subagent runs on the wrong workflow. Treat it as the front door to a papers-aware research stack, not a single-shot summarizer.156installs