Validate the replication package for the sewage-house-prices project. Runs 10 checks covering script execution, file integrity, output freshness, dependency verification, data provenance, and README completeness (AEA format). This skill should be used when asked to "audit replication", "check the package", or "verify reproducibility".
SkillScore73Maintained
Automatic
◆Research pack
Last verified 8/8/2026
What it changes
Passed 13 of 18 test cases with the skill, and 9 of 18 without it.
Without skill50%
With skill73%
Checks it passed that the model alone missed
The report is saved to output/log/ replication_audit_[date].md.
The summary report table contains exactly 10 distinct check rows.
The output file timestamp must be strictly newer than its generating source script modification time.
and 2 more in the full report
5 fixed11 passed either way6 still failing0 broke
Pass-rate lift
+23 pp
Tokens
+11%
Agent turns
0%
Response time
-55%
Model gemini-3.6-flashMeasured 8/8/2026v116 of 22 cases passed with skill · 18 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
Please perform a full replication verification without specifying any argument path. Where should the verification start looking for files?
What was checked, without and with the skill
Checked
Without skill
With skill
The audit defaults to checking the Replication directory if present before falling back to project root.
Missed
Passed
case-04 · gemini-3.6-flash · 8/8/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md94 linessingle file
Audit Replication Package
Run end-to-end validation of the replication package for the sewage-house-prices project.
Input:$ARGUMENTS — directory containing the replication package (Replication/ by default), or here for the current project root.
Project-Specific Context
Expected Structure
Pipeline scripts: scripts/R/01_data_ingestion/ through scripts/R/06_analysis_datasets/
Analysis scripts: scripts/R/09_analysis/
Utilities: scripts/R/utils/
Python: scripts/python/
Data: data/raw/, data/processed/, data/final/
Output: output/tables/, output/figures/
Manuscript: docs/overleaf/
R packages: renv.lock
Workflow
Step 1: Locate Package
If $ARGUMENTS is a directory path: use it
If $ARGUMENTS is here or empty: use Replication/ if it exists, otherwise project root
Verify the directory exists and contains scripts
Step 2: Run 10 Verification Checks
Standard Checks (1-4)
Script syntax — All .R files parse without syntax errors
File references — All read_parquet(), readRDS(), read.csv() targets exist or are documented as inputs
File integrity — No orphaned scripts (referenced but missing), no orphaned outputs (generated but unreferenced)
Output freshness — Output files are newer than their source scripts (no stale outputs)