A paper is not a list of experiments. It is an argument that a reader can follow, object to, and eventually believe. This skill helps turn research material into that argument before drafting prose.
Use this when
- The user has an idea, result set, outline, abstract, or rough draft and asks how to make it into a paper.
- The contribution is unclear or too broad.
- Experiments exist but the story does not.
- The user needs to decide what each section should prove.
- The paper has too many claims, disconnected tables, or weak motivation.
Do not use this for
- Polishing finished prose sentence by sentence. Use a writing/editing skill instead.
- Writing a source-grounded literature survey. Use
literature-triangulationorrelated-work-writer. - Checking whether results are reproducible. Use
reproducibility-audit. - Strategic rebuttal to reviewers. Use
reviewer-response-strategist.
Workflow
1. Extract the raw material
Identify, without embellishing:
- Problem: what is hard, missing, or misunderstood?
- Proposed idea or method: what did the authors do?
- Evidence: what results, proofs, ablations, or analyses exist?
- Audience: which community must be convinced?
- Constraint: venue, page limit, deadline, available experiments.
If these are missing, ask targeted questions instead of inventing them.
2. State the argument spine
Write the paper's argument in this form:
Because [problem/gap], we need [capability or understanding].
Existing work fails because [specific limitation].
This paper contributes [method/claim/analysis].
The key evidence is [main result + supporting result + stress test].
Therefore readers should believe [bounded conclusion].
The spine should be narrow enough that one paper can actually support it.
3. Separate claims by strength
Classify each claim:
- Main claim — the paper lives or dies on this.
- Supporting claim — needed to make the main claim believable.
- Mechanism claim — explains why the method works.
- Scope claim — where it works and where it does not.
- Speculative claim — interesting but not established; move to discussion or remove.
For each claim, require evidence. If evidence is missing, mark it explicitly.
4. Build the section logic
Create a section plan where every section has a job:
- Abstract — compressed problem, contribution, evidence, implication.
- Introduction — why the problem matters, what gap exists, what this paper proves.
- Related Work — how prior work creates the gap or contrast.
- Method / Theory — what is new enough to understand the contribution.
- Experiments / Results — evidence for each claim, in order of importance.
- Analysis / Ablations — why the result happens and what alternatives fail.
- Limitations — bounded honesty, not self-sabotage.
- Conclusion — what changed after reading the paper.
Do not include a section just because papers usually have one. Include it because it performs a necessary role in the argument.
5. Plan figures and tables as evidence
For each proposed figure or table, specify:
- claim it supports
- expected reader takeaway
- required comparison or control
- whether it belongs in main paper or appendix
Decorative figures are liabilities. If a figure does not support a claim, cut it or move it to appendix.
6. Stress-test reader objections
List the strongest likely objections:
- novelty objection
- baseline fairness objection
- generality/scope objection
- mechanism objection
- missing ablation objection
- reproducibility objection
For each, say whether the current paper answers it with evidence, prose, or neither.
Output format
Use this structure:
# Paper Argument Plan
## One-sentence thesis
## Argument spine
## Claim hierarchy
| Claim | Type | Evidence | Status |
|---|---|---|---|
## Section plan
## Figure/table plan
| Artifact | Claim supported | Takeaway | Main/appendix |
|---|---|---|---|
## Weak spots
## Next actions
Quality bar
A good plan should make the paper easier to reject or accept. If the argument is vague enough that no experiment could falsify it, sharpen it before writing prose.