Replication Package: A Landscape of Agentic AI Engineering — An Empirical Analysis of Job Postings

This replication package contains all data, coding artifacts, inter-rater reliability assessments, and statistical analyses for reproducing the study findings. The study analyses 210 agentic AI job postings collected from four platforms (Glassdoor, LinkedIn, SimplyHired, Indeed) across 11 countries and across 6 continents using Socio-Technical Grounded Theory for Data Analysis (STGT4DA). Contents:- Raw data from all 4 platforms (3,009 postings), all the links and descriptions of the postings are also available in the respective sheet for reproducing the results. - Final analysis corpus (210 postings) with full descriptions and metadata- Complete codebook with raw-to-canonical code mappings- Taxonomies: 34 responsibility codes, 45 hard skill codes, 18 soft skill subcategories- Binary matrices (responsibility × job, hard skill × job, soft skill × job)- Inter-rater reliability: coding phase (κ = 0.76, 30 postings, 718 statements) and filtering phase (κ = 0.763, 60 postings)- Second coder data with all agreements and disagreement resolutions- Co-occurrence analysis: Mann-Whitney U tests with Benjamini-Hochberg FDR correction for all 44 hard skills All the data and analysis can be validated and used to reproduce the findings of the study. Full author details will be disclosed upon acceptance.