Pre-Deployment Verification of LLM-Generated Robot Artifacts: A Systematic Review — Data and Code Archive

Data and code archive for the systematic review "Pre-Deployment Verification of LLM-Generated Robot Artifacts: A Systematic Review" (single author; manuscript under review). Contents:- Screening and extraction data: PRISMA 2020 flow counts, screening decisions, and the verified extraction layer for all assessed studies (71 initially included; 67 in the final corpus after the A1 second screen).- Coding instrument: frozen coding rubric (guarantee decision procedure Q1–Q5, family definitions, integration test) and the coding accuracy audit records (ground-truth re-coding of all 71 studies with per-study rationale).- Search documentation: versioned search strings for Scopus, Web of Science, IEEE Xplore, and arXiv (PRISMA-S).- Scripts: table/bibliography generators (build_bib.py), numeric consistency checks (check_counts.py), reliability computation (Cohen's kappa with bootstrap CI), and figure generation.- MANIFEST with SHA-256 hashes of all archived files. The review maps pre-deployment verification methods for LLM-generated robot code, task plans, behavior trees, and temporal-logic specifications (final corpus n = 67, from 71 initially included; 2022 to 2026).