1# Contract (F6-conf-04)2**Question:** causal effect of one extra training hour per employee on next-year productivity growth (pp/hour).3**Estimand:** coefficient on training_hours in a linear, homogeneous-effect model; identified under selection on observables given skill_index (the only pre-programme covariate in the file). It's meaningful only if no other confounders exist, skill_index is measured before treatment (stated), and the functional form is right.4**Baseline:** OLS productivity_growth ~ 1 + training_hours (the unadjusted model).5**Outputs:** out/results.json (naive_estimate, naive_ci_low/high, baseline_metric, model_metric, effect_estimate, effect_se, effect_ci_low/high, skill_index_coefficient, conclusion, task_id, feasibility, feasibility_reason, data_issues, n_input_rows, n_analysed, random_seed); out/figure1.png; out/table1.csv (model, estimate, se, ci_low, ci_high); out/report.md (6 sections); out/analysis.py; out/claims.json.6**Information rules:** use only data/firms.csv; no imputation; skill_index is pre-programme, so it is a legitimate control. productivity_growth is the outcome and must never be used as a control.7**Invalidity:** unobserved confounding; post-treatment controls; a missing confounder making the estimand inestimable (premise_invalid). Here skill_index is available, so the task is feasible.8