Triple-negative breast cancer (TNBC) is an aggressive and molecularly heterogeneous breast cancer subtype. We developed a six-gene oxidative stress-related signature (BAG5, ANXA2, LONP1, GBE1, CBR1, and DDIT4) in GSE58812 (107 patients; 29 deaths) and evaluated its transportability in GSE21653 (83 TNBC patients; 25 disease-free-survival events). Apparent discovery performance was high (C-index 0.823; 1-, 3-, and 5-year AUCs 0.962, 0.838, and 0.871), but repeated 5-fold out-of-fold evaluation across 50 repeats yielded a lower mean C-index of 0.764, indicating optimism in the apparent estimates. In GSE21653, the fully frozen continuous score showed modest discrimination for disease-free survival (C-index 0.626). The discovery-derived cutoff classified all 83 external patients as low risk; therefore, the categorical classifier was not externally validated and should not be used clinically. Bulk-transcriptome analyses associated higher discovery-cohort risk with hypoxia-related and ferroptosis-protection gene-expression programs and with lower interferon-gamma, cytotoxic T/NK, and antigen-presentation signatures. CD276 and VTCN1 transcript levels were also higher in the high-risk group. These immune and redox findings are transcriptomic associations rather than direct measurements of immune-cell abundance, ferroptosis sensitivity, or checkpoint-protein expression. The signature and its biological associations are therefore hypothesis-generating and require validation in larger endpoint-matched cohorts and functional studies.