ai-paper-review
Skill0 starsChanged 30 days agoMITView on GitHub
Review an existing AI, ML, NLP, or LLM manuscript for argument, novelty framing, evidence alignment, terminology, readability, section responsibilities, page balance, and appendix organization. Use for manuscript critique or a revision plan without rewriting. Technical citation, numerical, build, and venue audits belong to ai-paper-check.
In this repository2 other files it gives its agents.
- ai-paper-checkSkill
- ai-paper-writingSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Read the shared principles.
- Distinctive: line 36
Use the findings contract, prioritizing consequential issues.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI Paper Review
Read the shared principles. Review the requested scope without editing the manuscript or author's plan.
Reconstruct and assess
Recover the problem, closest prior limitation, insight, contribution, evidence, and supported implication. Reuse an existing paper plan as a statement of intent, then test it against the manuscript. Distinguish a missing explanation from missing research.
- Contribution: identify the concrete difference from closest work and the support for each advertised claim. Do not certify novelty from an incomplete literature search.
- Coherence: find the first missing dependency, undefined concept, abrupt transition, or repeated argument. Explain its consequence for the reader.
- Meaning: flag misleading terminology, unsupported causal explanations, strengthened claims, and selectively reported results.
- Presentation: identify table narration, defensive filler, and details that obscure the central evidence. Preserve meaningful uncertainty and established author names.
Load a guide only for the section under review:
| Scope | Direct guidance |
| Title or abstract | Title and abstract |
| Gap and contribution structure | Introduction |
| Positioning | Related work |
| Task or motivation | Setup and observations |
| Mechanism or theory | Method |
| Evaluation and interpretation | Experiments |
| Closing claims | Conclusion and limitations |
| Allocation or supplement | Page budget, appendix |
Judge each section against its purpose and paper type. Editorial defaults are preferences, not scientific acceptance criteria. Verify material comparisons or terminology against authoritative sources when needed; keep unavailable verification unresolved.
Deliver findings
Use the findings contract, prioritizing consequential issues. For review-and-rewrite, carry the same findings to writing and implement authorized revisions. Use technical checks for necessary factual verification; a handoff does not mean verification passed.
Example: theory paper review
Synthetic input: “For convex, L-smooth functions, Theorem 1 proves the bound under independent samples.” The conclusion says “The bound holds for arbitrary objectives.” No experiments are supplied; review only is requested.
Confirmed / critical: conclusion → scope exceeds the theorem → Theorem 1 requires convexity, smoothness, and independent samples → the central guarantee is overstated → restore those conditions in the conclusion. Missing experiments and an absent Empirically paragraph are not defects by themselves. Leave source files unchanged and do not claim to have verified the proof.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

