agentleFS
Sign inStart free
JerryLiu369JerryLiu369/awesome-quant-skills/skills/backtesting-audit/paper-repro/SKILL.md

paper-repro

Skill3 starsChanged 33 days agoMITView on GitHub

Goal: independent verification, not transcription. The reproduction must be able to disagree with the paper. Before touching code, write down from the paper: 1. Headline claims with exact numbers (table and page) 2. Data sources, date ranges, universe definitions 3. Every specification in every reported model (they will matter later) 4. What would count as a failed reproduction If any of these are ambiguous, list the ambiguities first; ambiguity resolution choices explain many "failed" reproductions. Implement the SINGLE headline result…

3 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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
  1. Distinctive:

    Before touching code, write down from the paper:

    line 18
  2. Distinctive:

    Every specification in every reported model (they will matter later)

    line 21
  3. Distinctive:

    If any of these are ambiguous, list the ambiguities first.

    line 24
  4. Distinctive:

    Prefer writing your own implementation over running authors' code first.

    line 30

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
<!-- Imported from: https://github.com/saksham10arora-dotcom/quant-skills Source Path: skills/paper-repro/SKILL.md Curated for awesome-quant-skills -->

name: paper-repro description: Structured workflow for reproducing a quantitative research paper or blog post analysis. Use when the user wants to reproduce, replicate, verify, or re-implement results from a paper, or asks if a paper's findings hold up.

Paper Reproduction Workflow

Goal: independent verification, not transcription. The reproduction must be able to disagree with the paper.

Phase 1: Extract the claim contract

Before touching code, write down from the paper:

  1. Headline claims with exact numbers (table and page)
  2. Data sources, date ranges, universe definitions
  3. Every specification in every reported model (they will matter later)
  4. What would count as a failed reproduction

If any of these are ambiguous, list the ambiguities first; ambiguity resolution choices explain many "failed" reproductions.

Phase 2: Minimum viable replication

Implement the SINGLE headline result only. Resist implementing everything.

  • Match data vintage where possible; document every forced substitution
  • Prefer writing your own implementation over running authors' code first; then diff against their repo if available
  • Log every deviation as you make it: DEVIATION: used X because Y

Phase 3: Verdict taxonomy

Classify the outcome honestly:

  • REPLICATED: numbers match within rounding
  • REPLICATED WITH DEVIATIONS: same conclusion after documented substitutions
  • PARTIALLY REPLICATED: some tables/robustness checks hold, others do not; say which
  • NOT REPLICATED: cannot recover headline numbers; report your best-guess numbers alongside theirs
  • NOT REPRODUCIBLE: impossible to attempt (no data, no spec detail); list exactly what is missing

Phase 4: Robustness the paper skipped

Run at minimum:

  • Alternative window/subsample
  • One specification perturbation the authors did not report
  • Sensitivity to the biggest discretionary choice you found in Phase 1

Output format

PAPER: <title, year>
CLAIM TESTED: <exact claim>
VERDICT: REPLICATED | ... (taxonomy above)
MY NUMBERS vs PAPER: <side by side table>
DEVIATIONS: [numbered list]
ROBUSTNESS: [what you perturbed, what happened]
ARTIFACTS: <files/scripts produced>

Save the verdict block: failed and partial reproductions are publishable knowledge too.

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.

Check your setup, free