Academic Research Journal

Academic Research Journal

Core Rule

Evaluate and build academic work as a chain of claims: problem -> literature gap -> question -> design -> data -> measurement -> analysis -> interpretation -> contribution. Do not reward polish when the research logic is weak, and do not dismiss useful work merely because realistic constraints create limitations.

First Pass

1. Classify the task: evaluate, design, draft, revise, review, respond to reviewers, audit sources, or plan evidence synthesis.
2. Classify the article type: quantitative, qualitative, mixed methods, experimental, quasi-experimental, survey, field, case study, action research, policy/program evaluation, computational/big-data, literature review, systematic review, meta-analysis, theory, commentary, critique, or non-academic essay.
3. Extract the research skeleton: problem, purpose, question, theory, population/case, sample/data, measures, design, analysis, findings, limitations, contribution, and claim strength.
4. Identify the intended inference: exploratory insight, description, association, causality, mechanism, prediction, interpretation, policy/practice decision, or cumulative evidence.
5. Define the evidence needed and mark gaps as I/I = insufficient information; mark inapplicable criteria as N/A.

Required Reads By Task

• Existing paper, essay, or article evaluation: tasks/evaluate-existing-work.md, references/research-quality-principles.md, and references/journal-readiness-scorecard.md.
• Article type or methods classification: references/article-type-matrix.md.
• Section-level critique: references/section-evaluation-matrix.md.
• Sampling, measures, experiments, qualitative, quantitative, mixed methods, survey, or program evaluation: references/methods-evaluation-matrix.md.
• Literature review, systematic review, or meta-analysis: tasks/evidence-synthesis.md and references/evidence-synthesis-and-review.md.
• Drafting a manuscript, essay, thesis chapter, or journal article: tasks/manuscript-production.md.
• Research design or empirical validation planning: tasks/design-research-project.md.
• Reviewer simulation, revision strategy, or response letter: tasks/revision-and-peer-review.md.
• Reference, citation, source, ethics, or integrity audit: tasks/source-and-reference-audit.md and references/integrity-and-source-audit.md.

Evaluation Protocol

Use the 1-5 scale unless the user requests another format: 5 strong; 4 mostly strong; 3 adequate but limited; 2 major weaknesses; 1 invalid or not journal-ready; N/A; I/I.

Always evaluate fit between the claim and the method:

• Generalization requires a defensible population, sampling frame, recruitment logic, response/nonresponse handling, and subgroup size.
• Causal language requires random assignment, credible quasi-experimental logic, temporal ordering, comparison conditions, and confound control.
• Measurement claims require conceptual definitions, operational alignment, reliability, validity, and bias mitigation.
• Qualitative claims require design fit, sampling rationale, recruitment detail, coding transparency, reflexivity, context, triangulation or other credibility checks, and analyzed evidence.
• Mixed methods claims require explicit design, clear strand roles, integration, value added, and handling of contradictory findings.
• Evidence-synthesis claims require transparent search, inclusion/exclusion criteria, study quality/bias assessment, heterogeneity handling, and interpretable synthesis.

Evidence And Verification

• Never invent sources, quotations, statistics, DOI links, journal facts, methods, datasets, reviewer comments, or results.
• Current facts such as journal scope, instructions, rankings, indexing, impact factor, citation counts, reporting standards, and software versions need live verification.
• Treat statistical significance as limited evidence; check magnitude, practical importance, sampling assumptions, model choice, robustness, and replication.
• Treat missing methods detail as an evaluation finding, not permission to infer.
• Separate evidence, inference, speculation, and recommendation.

Script Helper

• Run scripts/manuscript_static_audit.py for a heuristic scan of section coverage, overclaiming, missing limitations/ethics cues, significance-language risks, and reference signals.

Output Standard

Lead with the highest-impact judgment. Name assumptions, article type, intended inference, strongest contribution, fatal or major weaknesses, concrete fixes, and residual uncertainty. Use scorecards and claim-audit tables for evaluations; use manuscript architecture and section builders for drafting.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: oghie
• Source: oghie/skillsets
• License: MIT

Install and usage instructions live in the source repository linked above.