From jeem-skills
Assembles data/code deposit for JEEM manuscript replication: documents provenance of monitoring, satellite, and proprietary environmental data, and writes a README that reproduces every exhibit.
How this skill is triggered — by the user, by Claude, or both
Slash command
/jeem-skills:jeem-replication-packageThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- The paper is heading to submission or acceptance and the data/code deposit is not assembled
JEEM has a long-standing policy: all data used must be clearly documented, computational methods explained in enough detail to permit replication, and the data made available to any researcher for replication — and the Editor must be notified at submission if these conditions cannot be met (检索于 2026-06;以官网为准; re-check the current guide for authors, as Elsevier data-availability and code-sharing requirements are tightening). Environmental papers routinely hit this wall because the data are spatial, sensor-derived, or proprietary; the package must show a credible reproduction path even when raw data cannot be posted.
Three structural features make environmental replications fragile, and the package must address each head-on. First, versioning: satellite and weather products are reissued (collection upgrades, reanalysis vintages), so a result is irreproducible unless the exact version is pinned. Second, spatial operations: geocoding, reprojection, and spatial joins are easy to do interactively and impossible to reproduce unless scripted with the CRS recorded. Third, proprietary inputs: utility, parcel, and licensed sensor data often cannot be posted at all, so the package must substitute derived/synthetic data and a precise access protocol. A package that ignores these three will pass a casual check and fail a serious one.
make / run_all) that goes raw → cleaned → every table and figure, with no manual steps. Map each exhibit to the script that builds it.JEEM's policy requires telling the Editor at submission if the data-availability conditions cannot be fully met — do not wait for acceptance. Draft a short data-availability statement that says, per source, whether it is bundled, derived-and-bundled, or restricted-with-access-protocol, and name any source you genuinely cannot share with the reason (license, confidentiality). Pair every restricted source with what you can provide: aggregated data, the code that processed it, and synthetic/sample data so the pipeline runs. A candid, complete statement at submission is far better received than a discovered gap at the replication check.
JEEM replications break most often on the data layer, because environmental inputs are heterogeneous and versioned. Name the exact source and vintage for each:
A paper merges satellite PM2.5, EPA monitors, and assessor parcel data to estimate amenity capitalization. The replicator cannot reproduce the exposure variable because the satellite product version and the cloud mask were never stated, and the parcel-to-grid spatial join was done by hand. The JEEM fix: a master script that pulls the named satellite collection version, applies the recorded quality mask, and runs the spatial join with the logged CRS; the proprietary parcel data is replaced with a synthetic parcel file so the code runs end-to-end, plus the exact assessor access protocol for the real data. The package now reproduces every exhibit from one command.
Treat the package as part of the paper, not a chore at the end. Environmental-economics estimates feed policy debates, so a result that cannot be independently reproduced is a result a regulator should not act on — and JEEM's policy reflects that. Building the master script and provenance documentation as you go (rather than reconstructing them at acceptance) also catches your own errors early: a pipeline that runs clean from raw data to every exhibit is itself evidence of correctness. The packages that fail are the ones assembled hastily after the science is "done"; the ones that pass were designed for reproduction from the first analysis script.
Write the README for someone who has the package and none of your tacit knowledge. It should state, up front: the software and versions, the single command that reproduces everything, the expected runtime and memory (large raster/grid jobs can run for hours), and a table mapping each manuscript exhibit (Table 1, Figure 3, ...) to the script and the output file that produces it. Then a data section listing every source with its provenance block, flagging which sources are bundled, which are derived, and which require an access protocol. A replicator should be able to reproduce every shareable result without emailing you — and know exactly how to obtain the restricted pieces.
【Journal】Journal of Environmental Economics and Management
【Skill】jeem-replication-package
【Master script】one entry point reproduces all exhibits? [Y/N]
【Data provenance】every source: version/access-date/license/unit/CRS? [Y/N]
【Restricted data】derived data + access protocol + sample data provided? [Y/N]
【Editor notice】notified at submission if conditions unmet? [Y/N / n/a]
【Spatial pipeline】geocode/projection/buffers/joins scripted with CRS? [Y/N]
【Environment】software/package versions + seeds pinned? [Y/N]
【Source status】verified URL / 待核实 / not asserted
【Next skill】jeem-referee-strategy
npx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin jeem-skillsBuilds a data and code replication package for an Economic Journal manuscript to RES/EJ Data Editor standard, including README, data documentation, and Zenodo deposit preparation. Does not run the analysis.
Assembles the data-and-code replication package for a REStat manuscript, including deposit to the REStat Harvard Dataverse with a replication-permitting README.
Assembles a data and code replication package for a JPE manuscript to DCAS/JPE Data Editor standards, including README, data, code, and output, for deposit to the JPE Dataverse.