agentsclimarketplace

Read theory

Skill QinghongLin/paperdoctor/skills/read-theory

PaperDoctor: Evidence-Grounded and Actionable Feedback for Scientific Papers in Progress

Install
npx -y skills add QinghongLin/paperdoctor --skill read-theory

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 3 stars3 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

Verify theoretical claims — mathematical derivations, proofs, and formal arguments. Targets claims with evidence_type "theoretical" from check_claim.json.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.9 KB, as published. Nobody here has run it

Verify Theory

Check each claim tagged evidence_type: "theoretical" by reading the paper's arguments and verifying them step by step.

Workflow

- [ ] Step 1: Load claims
- [ ] Step 2: Verify each argument
- [ ] Step 3: Save report

Step 1: Load Claims

Read {paper_dir}/reports/check_claim.json. Filter to claims where evidence_type includes "theoretical".

Each claim's id, source, quote, claim, evidence_type fields should be copied as-is into the output — do not modify them.

Read the full paper text from {paper_dir}/metadata/{arxiv_id}/mathpix/{arxiv_id}.md.


Step 2: Verify Each Argument

For each theoretical claim, locate the relevant reasoning in the paper. Check:

  • Correctness — does each step follow from the previous one? (algebraic, logical, or conceptual)
  • Assumptions — are any implicit assumptions missing or unjustified?
  • Edge cases — does the result hold at boundary conditions?
  • Notation consistency — are symbols used consistently throughout?

For each claim, produce:

{
  "id": 3,
  "source": "method",
  "quote": "Eqn. 4 → 6, Section 3.1\n\nBy substituting Eqn. 4 into Eqn. 5, we obtain the closed-form loss in Eqn. 6",
  "claim": "The soft-label loss simplifies to a closed-form KL divergence",
  "evidence_type": ["theoretical"],
  "status": "pass",
  "reason": ""
}

Field definitions (paper triple → fields):

FieldWhenPurpose
quotealwaysThe Where evidence. Start with where in the paper the argument lives (equation range, section, theorem), then a blank line, then the original sentence under scrutiny.
reasonwarning/error onlyThe Why. What issue was found (e.g. "implicit assumption that distributions share support"). For pass you may include a one-line note about the verification work performed (e.g. "re-derived substitution from Eqn. 4 to 6").
suggestwarning/error onlyThe How. One-sentence concrete fix the author can apply.

Statuses:

StatusMeaning
passArgument is correct
warningCorrect but has implicit assumptions, ambiguity, or missing edge-case discussion
errorError found — wrong sign, dropped term, invalid logical step

When status is warning or error, reason should explain the problem precisely and suggest should propose a concrete fix.


Step 3: Save Report

Output: {paper_dir}/reports/check_theory.json

{
  "summary": {
    "total": 3,
    "pass": 2,
    "warning": 1,
    "error": 0
  },
  "results": [
    {
      "id": 3,
      "source": "method",
      "quote": "Eqn. 4 → 6, Section 3.1\n\nBy substituting Eqn. 4 into Eqn. 5, we obtain the closed-form loss in Eqn. 6",
      "claim": "The soft-label loss simplifies to a closed-form KL divergence",
      "evidence_type": ["theoretical"],
      "status": "warning",
      "reason": "Re-derived substitution from Eqn. 4 to 6 — derivation assumes teacher and student distributions share the same support, but this is not stated.",
      "suggest": "State the shared-support assumption explicitly before Eqn. 4."
    }
  ]
}

Coverage Rule

Every claim with theoretical in evidence_type MUST appear in results. No claim may be silently skipped. If a claim's theory is trivial or self-evident, still include it with status: "pass" and a brief note.

Tips

  • Reproduce each derivation from scratch — do not just skim the steps
  • For math: pay attention to summation indices and what gets absorbed vs cancelled
  • For non-math arguments: check if the logic actually supports the conclusion
  • A missing qualifier is a warning, not an error

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.