model-cardslisted

# Mitchell-Extended Model Cards This plugin produces 10-section model cards: Mitchell et al.'s 9 canonical sections plus an "Operational Details" 10th section for the consumer-evaluator audience. This skill is the reference for what each section is for and how to fill it honestly. ## The ten sections ### 1. Model Details What the model IS — name, provider, family, release date, parameter count where disclosed. Tier 1 (provider docs) is primary; tier 2 (HuggingFace) is secondary for open-source / fine-tuned models. ### 2. Intended Use What the model is for — primary uses, intended users, explicitly out-of-scope uses. The provider's documentation usually makes this explicit; quote it faithfully and cite. ### 3. Factors Subgroups, environmental factors, instrumentation. Often sparse for proprietary API-served models — write "Not publicly available" rather than fabricate. ### 4. Metrics Performance measures the provider reports. Distinguish provider-reported benchmarks from independent evaluations (tier 1 vs tier 4). ### 5. Evaluation Data Datasets used to evaluate. Frequently "Not publicly available" for proprietary models. Tier 3 (release paper) is the most likely source. ### 6. Training Data Datasets used to train. Almost always "Not publicly available" for proprietary models — say so. Tier 3 is the only source likely to disclose. ### 7. Quantitative Analyses Disaggregated performance — by demographic, by task category, by language. Tier 3 (release paper) is pr