Academic Research Skills (ARS)¶
external · status: active · focus: end-to-end · discipline: general · started: 2026
Project page: https://github.com/Imbad0202/academic-research-skills
Source: projects/landscape/academic-research-skills.yml
Positioning¶
A comprehensive Claude Code plugin suite (v3.9.0 at scoring date) for the academic research pipeline: literature → write → review → revise → finalize. Three sub-suites — Deep Research (13 agents), Academic Paper (12 agents), Academic Paper Reviewer (7 agents) — with Style Calibration, anti-leakage protocols, Semantic Scholar API verification, VLM figure verification, and explicit human-in- the-loop gates throughout.
Distinctive contribution¶
Most-adopted Claude Code-native research plugin (9k+ stars) with an explicit "AI as copilot, not pilot" design stance. v3.8 introduces a claim-faithfulness audit pass (ARS_CLAIM_AUDIT=1) that fetches each cited source against three-layer citation anchors and gate-blocks output on five hallucination classes (claim-not-supported, negative-constraint-violation, fabricated-reference, anchorless, constraint-violation-uncited). This is the most operationally serious anti-hallucination apparatus in the catalog.
Evaluation scores¶
| Dimension | Score (0–3) | Note |
|---|---|---|
| Lifecycle coverage | 2 | Five stages spanning literature through review; less coverage on empirical analysis / modeling than full-pipeline economics tools. |
| Autonomy level | 1 | Explicit copilot stance: 'AI is your copilot, not the pilot.' Human approval at every stage gate. |
| Architectural transparency | 3 | Detailed architecture docs (docs/ARCHITECTURE.md), public failure-mode reference, calibration thresholds (FNR<0.15, FPR<0.10) published in spec. |
| Inputs supported | 3 | Topic / draft / reviewer-comment inputs; Pandoc + tectonic for DOCX/PDF; multiple installation paths (plugin, project skills, global skills, claude.ai Project). |
| Outputs / reproducibility | 3 | Three-layer citation anchors enable claim-level audits; replication packages; versioned releases. |
| Internal evaluation | 3 | Calibration mode with gold-set FNR/FPR measurement; cross-model verification (ARS_CROSS_MODEL); engages published failure-mode literature directly. |
| Openness | 2 | CC BY-NC 4.0 (non-commercial); sponsor-supported; Codex CLI sibling distribution available. |
| Maturity / traction | 3 | 9.3k+ stars; v3.9.0 with multi-version release cadence; English + Traditional Chinese docs; multi-IDE support. |
| Cross-family policy | 1 | ARS_CROSS_MODEL flag enables cross-family verification; not the default. |
| Runtime assurance | 3 | Claim-faithfulness audit pass (ARS_CLAIM_AUDIT) with 5 HIGH-WARN classes + 3-stage citation anchors + cross-model verification = heaviest claim-audit harness in the catalog alongside ARIS. |
| Cross-platform portability | 3 | Claude Code + VS Code + JetBrains + Codex CLI sibling distribution + 5 install methods. |
Scored on 2026-05-18. See the evaluation rubric.
Tags¶
Pipeline stages: literature-discovery literature-synthesis paper-drafting revision-editing referee-simulation
Architectural features: multi-agent debate-consensus human-in-loop tool-use rag-knowledge-base artifact-versioning
Inputs: research-topic paper-draft reviewer-comments
Outputs: paper-pdf paper-docx literature-review reviewer-report revision-plan
Data sources: user-provided
Knowledge sources: semantic-scholar arxiv
Limitations¶
- Non-commercial license restricts deployment options.
- Heavy on the writing/review block; lighter on empirical data analysis or formal modeling.
- Quality of claim-audit gating depends on user-supplied calibration set.
- Author's substantive scholarly background is not disclosed in repo metadata; project is engineering-led rather than discipline-led.
Related projects in this catalog¶
Also compared in¶
- ARIS Table 4 (footnote) (
yang2026aris) — Discussed as a parallel claim-audit project.
Related references (literature catalog)¶
- Ji, Z. et al. (2023). Survey of Hallucination in Natural Language Generation
ji2023hallucination - Schick, T. et al. (2023). Toolformer: Language Models Can Teach Themselves to Use Tools
schick2023toolformer - Wu, J. et al. (2025). Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools
wu2025agenticreasoning