agentleFS
ShaishavMaisuriaShaishavMaisuria/research-paper-lifecycle-skills/skills/reflect-paper/SKILL.md

reflect-paper

Skill59 starsChanged 5 months agoApache-2.0View on GitHub

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.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run a structured reflection loop over a research-paper draft.

    line 16
  2. Distinctive:

    If the runtime supports delegated agents or subagents, split the reflection into two independent workers:

    line 27
  3. Distinctive:

    If delegation is unavailable, run the same passes sequentially.

    line 37
  4. Very specific:

    If a venue or CFP is involved, use parse-cfp or add-venue-profile before relying on page limits, blind level, required sections, or …

    line 74
  5. Distinctive:

    Read for correctness and evidence.

    line 79
  6. Distinctive:

    Read for communication and persuasion.

    line 109
  7. Distinctive:

    Use this priority scale:

    line 145
  8. Distinctive:

    Never invent papers, results, deadlines, page limits, or venue rules.

    line 167
  9. Distinctive:

    Do not modify technical claims to sound stronger unless the evidence supports them.

    line 168
  10. Distinctive:

    Keep confidential reviews and unpublished paper content local.

    line 170

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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:

  1. Send a Researcher worker the draft, target venue/context, and the Researcher pass checklist only.
  2. Send a Writer worker the draft, target venue/context, and the Writer pass checklist only.
  3. 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:

StagePrimary concern
Early draftMissing contribution, weak framing, incomplete evidence.
Pre-submissionVenue fit, page budget, anonymization, citation reliability, reviewer risk.
RebuttalReview coverage, evidence anchors, tone, response budget.
Camera-readyRequired final-file steps, de-anonymization, acknowledgments, source consistency.
PresentationTalk 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:

NeedUse
Find missing related workfind-papers, literature-review
Fetch legal OA copiesfetch-paper
Check bibliographyverify-citations
Study venue exemplarsstudy-exemplars
Simulate reviewer risksimulate-reviewers
Check CFP/venue factsparse-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:

NeedUse
Abstract rewrite or lintwrite-abstract
Related work draftingdraft-related-work
Full review synthesisliterature-review
Prose polishpolish-prose
Tables and figurespolish-tables-figures
Talk/poster translationmake-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:

PriorityMeaning
P0Blocks submission or creates serious correctness/integrity risk.
P1Likely affects review outcome or reader understanding.
P2Improves clarity, polish, or confidence.
P3Nice-to-have cleanup.

5. Output

Return a concise reflection report:

  1. Verdict — ready / close / needs major revision / not ready.
  2. Researcher pass — top technical, evidence, citation, and venue risks.
  3. Writer pass — top framing, structure, prose, and visual risks.
  4. Merged action plan — prioritized table with owner/pass, location, action, next skill, and estimated effort.
  5. 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.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.