Co-Researcher
poemswe/co-researcherAcademic research plugin with 15 skills; literature reviews check citations and quotes against OpenAlex, arXiv and Europe PMC.
Recommended
Check first Depth 4/5
Academic Research Skills (ARS) Claude skills
Five very detailed Claude Code skills for the academic pipeline: deep research, paper writing, peer review, revision and systematic-review screening.
Our verdict Recommended with caveats
Expert-grade and heavily tested, but licensed CC BY-NC 4.0 (non-commercial) per the README, and the 50-plus KB skills use a lot of context.
What Academic Research Skills (ARS) does
Academic Research Skills covers the path from research question to submitted paper with five skills and a team of sub-agents. deep-research runs a 13-agent research team with a Socratic guided mode, PRISMA-style systematic review and source verification against Semantic Scholar and Crossref. academic-paper plans, drafts and formats a manuscript and keeps citations anchored to locators in the sources. academic-paper-reviewer simulates a review panel (editor, domain, methodology and perspective reviewers plus a devil's advocate) and writes a decision letter and revision roadmap. academic-pipeline chains the stages with integrity gates, a 'material passport' that carries provenance, and checkpoints where it waits for your confirmation. sr-screener handles study selection for systematic reviews and ships eight scripts.
The skills are unusually long: SKILL.md files average about 52 KB, each backed by 7 to 29 reference files, and the repo contains more than a thousand eval-related files and CI. An opt-in claim audit fetches each cited source and checks whether it supports the claim, and a write-scope guard needs a real Python install. The README says the human stays the author: it is built for human-in-the-loop use, not full automation, and it states plainly that it does not help hide AI use. Install through the Claude Code plugin marketplace or a symlink; there is a community wrapper for Pi.
Good for
- Researchers writing and revising papers who want citation verification and structured peer-review simulation
- Systematic reviews that need PRISMA-style screening and risk-of-bias steps
- Supervisors or authors who want integrity gates between pipeline stages
Not for
- Commercial use, since the README's license is CC BY-NC 4.0 (non-commercial)
- Quick one-off summaries, where the sheer size of the skills is overkill
- Fully unattended paper generation (it is designed to stop for your confirmation)
What's inside depth 5/5
Five skills averaging 52 KB with 7 to 29 reference files each, eight sr-screener scripts, a write-scope guard and over 1,100 eval-related files with CI.
- Skills
- 5
- Reference docs
- 94
- Scripts
- 8
- Tests / evals
- 1146
- Templates
- 52
- Avg SKILL.md
- 51 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- Claude Code (recent version for plugin packaging) and an Anthropic API key
- Optional: Python for the write-scope guard and some commands
- Optional: Pandoc or tectonic for DOCX and PDF output
- Optional: a second model for cross-model verification (ARS_CROSS_MODEL)
- Telemetry: none found; the verification steps call public scholarly APIs (Crossref, Semantic Scholar) with DOIs and titles, and cross-model checks send text to the second model you configure.
Security review
Review before installingA large set of academic research, paper-writing, peer-review and systematic-review screening skills, with local Python scripts for the screening workflow. The files available show no credentials sent anywhere, but bibliographic lookups send DOIs and titles to public scholarly APIs and optional cross-model checks send text to a second model you configure. Most of the repository, including the data-flow doc and a write-scope hook the text refers to, was not available to review.
Worth knowing
- Makes network requestsVerification steps call public scholarly services (Crossref, Semantic Scholar, arXiv, doi.org) with DOIs and titles; the README says optional cross-model checks send text to a second model you configure (consent-gated).deep-research/references/crossref_api_protocol.md:1
- Relies on a CLI or package we could not reviewOptional Pandoc and tectonic are used for DOCX/PDF output and are not part of the repository.README.md:59
What we checked
The repository (2864 files) on Oct 29, 2026. Not a full review: 2650 of 2864 files were not downloaded (large repository; we reviewed the skill folders); it relies on code outside the repository: about 2,650 of 2,864 repository files were not downloaded, including .claude-plugin manifests, docs/DATA_FLOWS.md, docs/SETUP.md, the scripts/ folder (pipeline-integrity checker, cross-model handoff) and the PreToolUse write-scope guard the skill text mentions; Crossref, Semantic Scholar and arXiv APIs and any second model configured for cross-model verification.
NVIDIA SkillSpector static scan: 115 pattern hits (Prompt Injection 66, Excessive Agency 14, Anti-Refusal 12, Memory Poisoning 9, YARA Match 6, Data Exfiltration 4). The hits are almost all the skills' own prompt-hygiene text: repeated 'instruction-data boundary' HTML comment blocks, 'ignore previous instructions' as a quoted example of what to treat as data, anti-refusal-style wording in reviewer personas, and 'print prompt' in a prompt-source file. The eight code files are local screening scripts and two workflow templates.
A review, not a certification. Skills change after we check them — read the files before installing.
License
README states CC BY-NC 4.0 — non-commercial use only; GitHub does not detect a standard license, so confirm with the author for any commercial use.
Can it run unattended?
Not really. The pipeline deliberately stops for your confirmation at every stage and integrity gate, so it is a supervised workflow, not an unattended one.
Alternatives to Academic Research Skills (ARS)
Deep Research Skill
199-biotechnologies/claude-deep-research-skillA single research skill with an 8-phase pipeline, parallel search, source scoring, claim and citation checks, and report output in Markdown, HTML and PDF.
With caveats
Check first Depth 4/5
ARIS (Auto-Research-In-Sleep)
wanshuiyin/Auto-claude-code-research-in-sleepA large markdown skill set for autonomous ML research: idea discovery, experiment runs and cross-model review loops where a second model critiques the first.
With caveats
Broad access Depth 4/5
Anti-Defensive Writing
Adkid-Zephyr/anti-defensive-writing-SkillA lightweight CN/EN academic-writing skill that restructures paper drafts around one strongest claim instead of hedged, evenly-weighted self-critique.
With caveats
No issues Depth 2/5
FAQ
What is Academic Research Skills (ARS)?
ARS is a set of five Claude Code skills for academic work: deep research, paper writing, simulated peer review, an end-to-end pipeline with integrity gates, and a systematic-review screener. It keeps the human as the author and verifies citations.
How do I install Academic Research Skills?
In Claude Code run /plugin marketplace add Imbad0202/academic-research-skills and /plugin install academic-research-skills, then try /ars-plan. The README also documents symlink and global installs and a Pi wrapper.
Can I use ARS commercially?
The README badge and text say CC BY-NC 4.0, which forbids commercial use. GitHub reports the license as unrecognized, so check the repo's LICENSE and ask the author if you need commercial terms.
Does ARS write a whole paper automatically?
It can draft a full paper, but the pipeline stops for your confirmation at each stage and runs integrity checks for hallucinated or unsupported citations. The README says you remain responsible for every claim.
Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.