PaperCourt v0.1.0

PaperCourt traces empirical machine-learning paper claims to exact sources, repeatable checks, and concrete falsifying experiments.

Included

• A self-contained papercourt skill and project installer.
• Decimal checks for values, differences, relative changes, ratios, and unweighted means.
• Explicit comparison controls with aligned, different, and unknown outcomes.
• Markdown, standalone HTML, and JSON reports with source locations, hashes, and byte-preserved inputs for replay.
• An original MIT-licensed fictional example covering correct claims, numeric conflicts, hardware confounding, overgeneralization, and missing evidence.

Install

npx skills@1.7.0 add K-kiron/PaperCourt --skill papercourt --agent codex

The report helper requires Python 3.10 or newer and has no third-party runtime dependencies. Extract papercourt-skill-0.1.0.zip into your project's .agents/skills/ directory for manual installation, or use the Python installer in the full source bundle.

Verification

• 28 behavioral tests pass.
• Windows and Linux CI pass on Python 3.10 and 3.12, including repository checks, example replay, and release construction.
• A fresh installation from the public GitHub repository completed the example audit.
• Independent builds produced identical archives; the extracted standalone skill reproduced the checked-in reports.

See the validation record and worked report.

Assets

• PaperCourt-0.1.0.zip: full source, documentation, tests, and examples.
• papercourt-skill-0.1.0.zip: standalone skill, helper, reference, metadata, and license.
• SHA256SUMS: checksums for both archives.
• report.html: standalone fictional demonstration; download and open locally.

Semantic assessments remain model analysis. Missing evidence is not a refutation. This release does not establish improvements in review quality, execute experiments, parse arbitrary PDFs, or issue academic verdicts. The helper makes no network calls; hosted agent processing follows the host's data policy.