Reflect paper
Skill ShaishavMaisuria/research-paper-lifecycle-skills/skills/reflect-paper
42 AI agent skills for literature review, academic writing, citation verification, conference submission, rebuttal, publication, and presentations.
npx -y skills add ShaishavMaisuria/research-paper-lifecycle-skills --skill reflect-paperAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 23 stars23 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Runs a two-pass paper reflection loop with a Researcher pass and a Writer pass. Use when the user asks to reflect on a paper, improve a draft, run a research-agent and writer-agent review, critique the paper before submission, find both technical and writing issues, or produce a revision plan. The Researcher pass checks claims, citations, evidence, venue fit, missing related work, and reviewer risk; the Writer pass checks framing, structure, abstract, related work, prose, tables/figures, and response clarity. Outputs a merged action plan with evidence, priority, owner, and next skill to run.
SKILL.md
6.8 KB, ~1.4k tokens by cl100k_base, as published. Nobody here has run it
Reflect Paper
Run a structured reflection loop over a research-paper draft. Treat this as two separate perspectives, then merge them into one revision plan:
- Researcher pass — technical truth, evidence, citations, venue fit, novelty, limitations, reviewer risk.
- Writer pass — narrative, framing, section structure, abstract, related work placement, prose, figures/tables, reader flow.
This is an orchestration skill: it calls on sibling skills when their narrower checks are needed, but it owns the reflection protocol and final synthesis.
If the runtime supports delegated agents or subagents, split the reflection into two independent workers:
- Send a Researcher worker the draft, target venue/context, and the Researcher pass checklist only.
- Send a Writer worker the draft, target venue/context, and the Writer pass checklist only.
- Merge both outputs yourself; do not let either worker decide the final verdict alone.
If delegation is unavailable, run the same passes sequentially. Keep separate notes for each pass so technical-risk findings do not flatten writing feedback, and writing polish does not hide correctness concerns.
When to use
- "Reflect on this paper before I submit"
- "Run a research agent and writer agent on my draft"
- "Critique this draft from both technical and writing angles"
- "Give me a revision plan"
- "What should I fix before sending this to my advisor / reviewers?"
- "Do a final paper reflection pass"
Inputs
- The draft: LaTeX source, PDF, markdown, Word-exported text, or pasted sections.
- Optional: target venue or CFP URL.
- Optional: bibliography file, related-work notes, reviews, or advisor feedback.
- Optional: the user's current goal, such as submission readiness, camera-ready polish, rebuttal, or presentation prep.
Process
1. Scope The Reflection
Identify the lifecycle stage and what "good" means for this pass:
| Stage | Primary concern |
|---|---|
| Early draft | Missing contribution, weak framing, incomplete evidence. |
| Pre-submission | Venue fit, page budget, anonymization, citation reliability, reviewer risk. |
| Rebuttal | Review coverage, evidence anchors, tone, response budget. |
| Camera-ready | Required final-file steps, de-anonymization, acknowledgments, source consistency. |
| Presentation | Talk story, timing, Q&A risk, slide/poster readability. |
If a venue or CFP is involved, use parse-cfp or add-venue-profile before
relying on page limits, blind level, required sections, or deadlines.
2. Researcher Pass
Read for correctness and evidence. Check:
- Core claim: what the paper says it contributes, and whether the evidence actually supports it.
- Novelty and related work: missing comparisons, closest competitors, thin or overstated deltas.
- Citations: unresolved entries, suspicious metadata, fabricated-looking references, uncited claims.
- Methods and evaluation: assumptions, baselines, ablations, datasets, reproducibility, statistics, limitations.
- Venue fit: track expectations, page budget, anonymity, required sections, LLM/AI-use policy, artifact expectations.
- Reviewer risk: likely objections and what evidence would defuse them.
Delegate as needed:
| Need | Use |
|---|---|
| Find missing related work | find-papers, literature-review |
| Fetch legal OA copies | fetch-paper |
| Check bibliography | verify-citations |
| Study venue exemplars | study-exemplars |
| Simulate reviewer risk | simulate-reviewers |
| Check CFP/venue facts | parse-cfp, select-venue, tailor-to-venue |
Record findings as: finding, evidence, risk, suggested fix, and
confidence.
3. Writer Pass
Read for communication and persuasion. Check:
- One-sentence paper promise: can a reader say what the paper does and why it matters?
- Abstract: motivation, gap, approach, result, impact, and venue-specific length/keyword rules.
- Introduction: problem setup, contribution bullets, reader contract.
- Related work: clustered themes and explicit deltas, not a laundry list.
- Method and evaluation: section order, figure/table references, claim titles, readable transitions.
- Prose: overclaiming, hedging, AI-sounding filler, terminology drift, passive contribution statements.
- Figures and tables: captions, crossrefs, color accessibility, layout, and whether each visual answers a reader question.
Delegate as needed:
| Need | Use |
|---|---|
| Abstract rewrite or lint | write-abstract |
| Related work drafting | draft-related-work |
| Full review synthesis | literature-review |
| Prose polish | polish-prose |
| Tables and figures | polish-tables-figures |
| Talk/poster translation | make-slides, write-talk-script, make-poster |
Record findings as: reader problem, location, revision, expected benefit, and next skill.
4. Reflection Merge
Merge the two passes into a prioritized plan. Do not average away conflicts: when the Researcher pass wants more caveats and the Writer pass wants a sharper claim, state the tradeoff and propose a wording that is both accurate and readable.
Use this priority scale:
| Priority | Meaning |
|---|---|
| P0 | Blocks submission or creates serious correctness/integrity risk. |
| P1 | Likely affects review outcome or reader understanding. |
| P2 | Improves clarity, polish, or confidence. |
| P3 | Nice-to-have cleanup. |
5. Output
Return a concise reflection report:
- Verdict — ready / close / needs major revision / not ready.
- Researcher pass — top technical, evidence, citation, and venue risks.
- Writer pass — top framing, structure, prose, and visual risks.
- Merged action plan — prioritized table with owner/pass, location, action, next skill, and estimated effort.
- Open questions — facts that require user input or live verification.
Guardrails
- Never invent papers, results, deadlines, page limits, or venue rules.
- Do not modify technical claims to sound stronger unless the evidence supports them.
- Keep confidential reviews and unpublished paper content local.
- For legal/copyright safety, fetch papers only from legal open-access sources and do not reproduce abstracts or paper text in the report.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most docs writing skills give in ~1.4k tokens
Counted across 1,637 of the 3,044 authors here whose files we hold, read 2026-08-07
- Announce the skill at startin 54 of 1637, across 26 files
- Convert legacy doc files before editingin 45 of 1637, across 7 files
- Predict questions readers might askin 42 of 1637, across 4 files
- Generate clarifying questions for initial contextin 42 of 1637, across 3 files
- Create document scaffold with placeholder textin 42 of 1637, across 3 files
- Brainstorm content options for each sectionin 42 of 1637, across 3 files
- Test the document with a fresh context-less instancein 42 of 1637, across 3 files
- Include exact file paths in every taskin 42 of 1637, across 15 files
- Ask interview questions one at a timein 42 of 1637, across 27 files
- Apply surgical edits during refinementin 41 of 1637, across 2 files
- Offer structured workflow or freeformin 40 of 1637, across 1 file
- Ask for document meta-contextin 40 of 1637, across 2 files
Said here and by no other author read
- merge the researcher and writer pass outputs yourself
- identify the draft lifecycle stage before reviewing
- record technical findings with evidence and confidence
- record writing findings with reader problem and revision
- return a concise reflection report
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.