Citation risk has a number.
Zhao et al. audited 111M references across 2.5M papers. The article uses the 146,932 estimate as the event hook.
The repo packages research, writing, review, revision, citation checks, and disclosure as 4 Claude Code skills. The article's claim: its strongest value is not autonomous paper writing, but mandatory handoffs and integrity gates around a known citation failure mode.
Zhao et al. audited 111M references across 2.5M papers. The article uses the 146,932 estimate as the event hook.
25 modes is the canonical count from the repo's mode registry. Plugin metadata claims "35+ modes" and diverges from the registry.
Claude Code users can add the marketplace, install the plugin, then start with /ars-plan.
The license is CC BY-NC 4.0. Commercial use needs separate permission from the maintainer.
Click a node to see what it owns. Solid lines are the main paper flow. Dashed lines are optional or observability paths.
The companion mirrors the article's install path. Checklist state is stored in this browser only and is not sent anywhere.
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skills
export ANTHROPIC_API_KEY=sk-ant-...
claude
/ars-plan
git clone https://github.com/Imbad0202/academic-research-skills.git ~/academic-research-skills
cd /path/to/your/project
mkdir -p .claude/skills
ln -s ~/academic-research-skills/deep-research .claude/skills/deep-research
ln -s ~/academic-research-skills/academic-paper .claude/skills/academic-paper
ln -s ~/academic-research-skills/academic-paper-reviewer .claude/skills/academic-paper-reviewer
ln -s ~/academic-research-skills/academic-pipeline .claude/skills/academic-pipeline
Three entry points cover most real usage. Pick the one that matches what you have on hand: an idea, a draft, or a reviewer report.
Walks all ten stages with checkpoint confirmation. Output is a finished APA 7.0 paper, an editorial decision letter, a revision roadmap, two integrity reports, and an AI self-reflection log.
/ars-full
Socratic chapter-by-chapter dialogue. Use it when the research question is not yet clear, when intent is exploratory, or when you want a chapter plan before drafting.
/ars-plan
Skip the orchestrator when only one function is needed. Eight slash commands cover citation checks, format conversion, disclosure statements, abstracts, outlines, and revisions.
/ars-citation-check + 7 more
Numbers come from the published article, the repo's release metadata, and GitHub API checks on 2026-05-13. Volatile GitHub counters use safe-floor ranges.
| Field | Value | Source note |
|---|---|---|
| Repo | Imbad0202/academic-research-skills | Public GitHub repo. |
| Release anchor | v3.7.0, published 2026-05-05 | GitHub latest release API. |
| Main branch work | v3.7.3 in progress, not a tagged release | Repo changelog and latest release check. |
| GitHub stars | +6.7k stars | As of 2026-05-13. Re-verify at read time, the velocity is high. |
| Registered modes | 25 | Canonical count from the repo's mode registry. |
| Skill folders | 4 | deep-research, academic-paper, academic-paper-reviewer, academic-pipeline. |
| License | CC BY-NC 4.0 | Source-available for noncommercial use, not OSI open source. |
| Command | Model | What it does |
|---|---|---|
/ars-full | opus | Runs the full research -> write -> review -> revise -> finalize workflow. |
/ars-revision-coach | opus | Parses reviewer comments into a roadmap and response-letter skeleton. |
/ars-plan | sonnet | Socratic chapter-by-chapter planning. |
/ars-outline | sonnet | Detailed outline plus evidence map. |
/ars-revision | sonnet | Revised draft plus response-to-reviewers. |
/ars-abstract | sonnet | Bilingual abstract plus keywords. |
/ars-lit-review | sonnet | Annotated bibliography in paper format. |
/ars-format-convert | sonnet | Converts between LaTeX, DOCX, PDF, or Markdown. |
/ars-citation-check | sonnet | Citation error report. |
/ars-disclosure | sonnet | Venue-specific AI-usage disclosure statement. |
| Repo | License | Pipeline scope | Plugin packaging | Citation-integrity surface |
|---|---|---|---|---|
Imbad0202/academic-research-skills | CC BY-NC 4.0 | Full 10-stage research, write, review, revise, finalize. | Claude Code plugin v3.7.0. | Stage 2.5 and 4.5 gates, Semantic Scholar verification, v3.7.3 Three-Layer Citation Emission in progress. |
aspi6246/Claude-Code-Skills-for-Academics | Verify before citing. | Read-only review focus. | Skill files only. | No equivalent multi-stage gate documented in this pass. |
Norman-bury/research-writing-skill | MIT | Cross-platform research writing assistant. | Claude Code, Cursor, Codex, OpenCode, Gemini. | No dedicated integrity-gate architecture documented in this pass. |
Galaxy-Dawn/claude-scholar | Verify before citing. | Project memory and toolchain integration. | Codex, OpenCode, Claude Code. | Toolchain-integration focus, not pipeline-integrity focus. |
Adoption constraints the article keeps visible. Each item is sourced to the article, the repo's release metadata, or the linked sources below.
Use "source-available" for the repo. Do not call it OSI open source unless the license changes.
A Codex sibling exists, but the article's install path and plugin assets target Claude Code.
The HTML is anchored to v3.7.0. Three-Layer Citation Emission should be described as in progress.
The draft says the showcase audit found 21 issues across 68 references after earlier automated checks.
The registry says 25 modes. Plugin metadata claims "35+ modes." The companion follows the registry.
Repo docs estimate $4 to $6 for a 15,000-word paper on Opus 4.7, plus optional cross-model cost.
Short answers drawn from the article.
Run /plugin marketplace add Imbad0202/academic-research-skills, then /plugin install academic-research-skills. The first practical command is /ars-plan.
No. The repo positions ARS as assistive, not autonomous. The pipeline has human checkpoints and two integrity gates.
Stage 2.5 and Stage 4.5 run integrity checks. v3.7.3 work on main adds citation locator markers after citation refs, but that work is not the tagged release.
The repo's performance docs estimate about $4 to $6 for a 15,000-word paper on Opus 4.7. Cross-model verification adds about $0.60 to $1.10.
The full article runs on X. Like or comment on the post to help more researchers find it, and to keep the conversation in one place.
Every claim in the companion is anchored to one of the linked sources below.
v3.7.0.