What this resource does
Core uses
About this resource
Academic Paper Skills contains a strategist Skill for venue analysis, literature gaps, and outline design, plus a composer Skill for drafting and checking philosophy or interdisciplinary papers from a detailed outline.
Both documented Skills are represented on this page.
Paper strategy and outline
Analyses the target venue, searches literature, examines research gaps, and develops a reviewer-assessed paper outline.
- Typical inputs
- Paper topic and research idea; Target venue and representative papers; User decisions at the documented checkpoints
- Output
- A writing guide, literature and gap reports, originality assessment, and optimised detailed outline.
Best for: Planning a philosophy or interdisciplinary paper before drafting.
Skill for this task
academic-paper-strategist
Plans the paper through venue, literature, gap, and outline stages.
Suggested requests (2)
Plan a paper on how mortality generates consciousness.
Identify research gaps in my topic and develop a detailed paper outline for the target preprint platform.
Adapted into complete requests from official trigger wording in strategist/SKILL.md.
Manuscript drafting and checks
Turns a detailed outline into chapter drafts and runs cross-chapter, completeness, and final quality checks.
- Typical inputs
- Detailed paper outline; Verified literature, citations, and writing requirements
- Output
- A manuscript draft, coherence report, completeness checklist, and final evaluation artifacts.
Best for: Drafting from an already approved and evidence-backed outline.
Skill for this task
academic-paper-composer
Drafts and checks a paper from a detailed outline.
Suggested requests (2)
Write the paper from this outline: [paste your detailed outline].
Compose the full manuscript from my approved outline and produce the documented quality reports.
Adapted into complete requests from official trigger wording in composer/SKILL.md.
Use boundaries
Limits and checks
Citation support
References or gap claims may be incomplete, incorrect, or unsupported.
Open and verify every cited source and check the argument against it.
Venue claims
A style guide or checklist may become outdated.
Compare the output with the current official venue instructions.
Official install gap
The README clone command contains a placeholder repository owner.
Open the official repository source, download or clone it there, then use the exact documented copy commands.
More boundaries
- Do not start the composer with only a vague topic; it requires a detailed outline and supporting literature.
- The Skills do not verify that every cited paper exists or supports the associated claim.
- A manuscript described by the source as submission-ready still needs author, supervisor, source, integrity, and venue checks.
Technical details
- Resource type
- Skill Collection
- Source last updated
- 4 Jan 2026
- Last verified
- 15 Jul 2026
- Documented AI models
- Claude
- Licence
- MIT
- Access
- Publicly available
- Additional costs
- Platform terms or usage limits may apply. API usage fees may apply for selected components. External services, software, compute, or data access may have separate costs.
- Skill instruction language
- English
- Documentation language
- English
- Repository languages
- Python, Markdown
- Dependencies
- Claude Code; Python 3.8 or later for verification scripts; Research sources and sample papers
- Review status
- Quick test completed
Continue exploring
Related resources
Official package contents
Files
Read the official English instructions without leaving this profile. This view includes the English files represented by this profile. Non-English source files, supporting code, assets, tests, and references may not be shown.
Official package files
2 files available here
composer1
strategist1
Read on this page
Select a file
Choose a file from the tree to read its official instructions without leaving this page.
Independent testing
Evaluation report
See exactly what was tested, which checks were used, and what the result does not cover.
Evaluations are mainly based on automated AI-assisted checks. Results are for reference only.
Evaluation coverage
2 components tested separately
Each score applies only to the named component, cases, application and model. The scores are not combined into a Repository rating.
Tested component 1academic-paper-strategist100 / 100 · One example
One example
Planning seven sections without overstating the main claim
The component returned a seven-section outline. It kept the claim tentative and did not turn the proposed method into a proven improvement in reasoning.
Codex CLI · gpt-5.6-luna · 4 Sept 2026
What we asked it to do
Turn a paper idea into a detailed outline. Keep its cautious claim, preserve the stated limitation and distribute a planned 5,400 words across seven sections.
Exact request used in the test
Create a detailed outline using the supplied claims, section order, citation relationships, and word budget. Do not strengthen the claims or introduce empirical findings.
What was provided
A synthetic paper idea about making philosophical arguments easier to inspect, a cautious central claim, a limitation, seven section budgets and one supplied citation relationship.
Original input
Title: Transparent Argument Maps for AI-Assisted Philosophy. Required claim: The method may make argument structure easier to inspect in some educational settings. Required limitation: The supplied evidence does not establish improved reasoning accuracy or any causal effect. Required section order and word counts: Abstract 200; Introduction 800; Related Work 900; Method 1,200; Illustrative Analysis 1,300; Limitations 600; Conclusion 400. Citation relationship: the Related Work section must associate Chalmers (1995), DOI 10.1017/S0140525X00032325, only with the hard problem of consciousness.
Output from the Skill
The complete original outline is below (774 whitespace-separated words). Numbers beside section headings are budgets for the future manuscript. They do not mean that those sections have already been written.
Transparent Argument Maps for AI-Assisted Philosophy
Abstract (200 words)
- State the motivating problem: AI-assisted philosophical work can be difficult to inspect when inferential structure is implicit.
- Present the bounded proposal: transparent argument maps as an inspectable representation of claims, reasons, objections, and dependencies.
- Preserve the qualified contribution: the method may make argument structure easier to inspect in some educational settings.
- Define the paper's scope as an illustrative analysis and methodological proposal, not an outcome study.
- State that the supplied evidence does not establish improved reasoning accuracy or any causal effect.
Introduction (800 words)
- Motivate inspection of philosophical arguments in educational settings.
- Distinguish visibility of structure from correctness, learning gains, or causal efficacy.
- Define AI-assisted philosophy only as the setting in which the proposed representation is considered.
- State the research question: how can an argument map represent inferential structure while preserving uncertainty and limitations?
- State the qualified thesis: the method may make argument structure easier to inspect in some educational settings.
- Preview the paper: conceptual motivation, related work, mapping procedure, illustrative analysis, and limitations.
- Clarify that no new empirical finding is claimed.
Related Work (900 words)
- Organize the discussion around argument representation, inspection, and educational use.
- Explain how argument maps can expose relations among claims, reasons, objections, and rebuttals.
- Separate representational transparency from epistemic validation.
- Discuss the supplied citation only in the context of the hard problem of consciousness: Chalmers (1995), DOI 10.1017/S0140525X00032325.
- Do not use that citation to support educational effectiveness, reasoning accuracy, AI capability, or causality.
- Identify the paper's narrow contribution as a preservation-oriented mapping framework and illustrative use, without adding sources or empirical results.
Method (1,200 words)
- Specify the unit of analysis: a claim and its explicitly linked reasons, objections, responses, and unresolved dependencies.
- Describe a mapping workflow: segment an argument; label propositions; connect support and challenge relations; mark uncertainty; record omitted premises as questions rather than facts.
- Define a transparency criterion: a reader can inspect the stated relation among mapped elements without treating the map as proof of correctness.
- Describe how AI assistance is bounded: generate or arrange candidate structure for inspection, while retaining human review of interpretation.
- Provide a representation schema using only supplied concepts: claim, reason, objection, response, dependency, uncertainty, and scope.
- Explain how an educational setting could use the map for inspection of structure, while withholding claims about learning or accuracy.
- Specify an illustrative comparison procedure: compare prose order with mapped relations and record where dependencies become explicit.
- State that this is a methodological outline and does not report an experiment, measured outcome, or causal estimate.
Illustrative Analysis (1,300 words)
- Introduce a bounded hypothetical philosophical argument as an illustration, without presenting it as empirical evidence.
- Show how a central claim is separated from its supporting reasons and possible objections.
- Mark which relations are explicit, which are interpretive, and which remain unresolved.
- Demonstrate how a reader might inspect whether a conclusion depends on a stated premise.
- Use the map to distinguish an argument's structure from the truth of its propositions.
- Explain how an educational reader could compare the map with the source prose.
- Discuss possible inspection benefits only in qualified terms: the method may make argument structure easier to inspect in some educational settings.
- Include a stopping rule: the illustration cannot establish improved reasoning accuracy or any causal effect.
- Avoid adding quotations, sources, bibliographic details, or empirical findings.
Limitations (600 words)
- State explicitly that the supplied evidence does not establish improved reasoning accuracy.
- State explicitly that the supplied evidence does not establish any causal effect.
- Explain risks of omission, misclassification, oversimplification, and apparently authoritative AI-generated links.
- Note that inspectability may vary across arguments, readers, and educational settings.
- Distinguish a proposed or illustrated method from a validated intervention.
- Identify evaluation requirements for future work without claiming that such work has been done.
Conclusion (400 words)
- Restate the bounded proposal: transparent argument maps can represent inferential structure for inspection.
- Preserve the modality and scope: the method may make argument structure easier to inspect in some educational settings.
- Reiterate the explicit limitation concerning reasoning accuracy and causality.
- Summarize the contribution as a qualified methodological design and illustrative analysis.
- End without adding empirical findings or expanding the citation relationship.
Allocation check
- Abstract: 200
- Introduction: 800
- Related Work: 900
- Method: 1,200
- Illustrative Analysis: 1,300
- Limitations: 600
- Conclusion: 400
- Total: 5,400 words
Our assessment
The word budget is a plan, not written prose
The seven allocations are 200, 800, 900, 1,200, 1,300, 600 and 400 words, totalling 5,400. The output supplies bullet-point guidance for writing those sections.
The central claim was not strengthened
The outline retains may and some educational settings. It explicitly says that improved reasoning accuracy and causal effects have not been established.
The outline adds suggestions, not research findings
It proposes a mapping procedure and hypothetical illustration. Those are ideas for the future paper, not measured results. The supplied citation remains confined to its requested topic and was not independently verified.
How this example was scored
This score covers one selected task using the existing five weighted dimensions. It is not a score for the whole component or repository. We reviewed the saved 4 September output; no new Skill execution took place. Unexercised checks are excluded, not counted as passes.
Five scoring dimensions
Passed checks receive 1 point, partly met checks 0.5, and failed checks 0. Checks outside this example are excluded, not counted as passes. The five dimension percentages are combined using the weights shown below.
Did it complete the task?
100.0%All seven outline sections and their word allocations are present.
Did it run without problems?
100.0%The saved prompt-only run completed and returned the outline.
Did it preserve and handle the supplied information responsibly?
100.0%The tentative claim and limitation remain explicit. Proposed method details are not presented as measured results.
Did it do the documented task in the tested setup?
100.0%Only this task in the recorded Codex setup is covered, not the complete component workflow.
Were the setup and use instructions clear enough to follow?
100.0%The guide describes the inputs and outline-design process used for this task.
Limits of this example
- One selected example is shown; the other nine earlier cases are not included in this score.
- Literature discovery, DOI verification, peer review, submission and the complete multi-stage workflow are not tested here.
- This was outline planning from supplied material, not a literature-grounded originality assessment or a finished paper.
Tested component 2academic-paper-composer96.3 / 100 · One example
One example
Turning supplied text and a table into a one-page PDF
The component produced a readable one-page PDF. The supplied text and table survived the formatting, including 35% and the statement that no reader outcomes were measured.
Codex CLI · gpt-5.6-luna · 4 Sept 2026
What we asked it to do
Put the supplied title, abstract, three sections, two-row comparison table and reference into a LaTeX document and PDF. Keep their content unchanged.
Exact request used in the test
Create a LaTeX manuscript and PDF containing the supplied title, abstract, numbered sections, table, limitation, and reference. Render every page and compare the generated file content with the supplied material.
What was provided
A short synthetic manuscript, not a real study: supplied sentences, a comparison of verbal and numeric confidence labels, and a reference to reproduce. No new research or writing from scratch was requested.
Original input
Title: A Bounded Comparison of Confidence Labels. Abstract: "This note compares verbal and numeric confidence labels without testing their effects." Section 1 heading: Definitions. Required sentence: "A verbal label uses words; a numeric label uses a percentage." Section 2 heading: Comparison. Table columns: Label type, Example, Claimed effect. Row 1: Verbal | low confidence | none tested. Row 2: Numeric | 35% | none tested. Section 3 heading: Limitation. Required sentence: "No reader outcomes were measured." Reference: David J. Chalmers. 1995. "Facing Up to the Problem of Consciousness." Behavioral and Brain Sciences 18(2): 200–219. https://doi.org/10.1017/S0140525X00032325.
Generated document
This is the actual generated PDF page, not the LaTeX code or a test log. The reference was supplied with the input; its bibliographic accuracy was not checked in this example.
What the result shows
All requested parts are visible
The PDF has the title, abstract, Definitions, Comparison, Limitation and reference. Both table rows fit on the page without overlapping text.
The values and limitation stayed intact
The numeric row still says 35%. Both rows say none tested, and the text still says that no reader outcomes were measured.
Formatting is not full-paper writing
The task provided the content in advance. This result does not demonstrate original argument development, a verified bibliography or completion of the full manuscript workflow.
How this example was scored
This score covers one selected task using the existing five weighted dimensions. It is not a score for the whole component or repository. We reviewed the saved 4 September output; no new Skill execution took place. Unexercised checks are excluded, not counted as passes.
Five scoring dimensions
Passed checks receive 1 point, partly met checks 0.5, and failed checks 0. Checks outside this example are excluded, not counted as passes. The five dimension percentages are combined using the weights shown below.
Did it complete the task?
100.0%The requested one-page document was produced with all supplied parts.
Did it run without problems?
100.0%The saved local compilation completed and produced a readable PDF.
Did it preserve and handle the supplied information responsibly?
100.0%The text, table values and limitation are preserved. Copying the reference is not verification of it.
Did it do the documented task in the tested setup?
100.0%Only this task in the recorded Codex setup is covered, not the complete component workflow.
Were the setup and use instructions clear enough to follow?
75.0%Inputs and output formats are described, but local compiler setup and commands are not supplied in the main guide.
Limits of this example
- One selected example is shown; the other nine earlier cases are not included in this score.
- Literature discovery, DOI verification, peer review, submission and the complete multi-stage workflow are not tested here.
- Only a short one-page document was formatted. Long manuscripts, complex figures and publisher templates were not tested.
Want to see more tests?
Tell us which module or task you would like us to test next.