Assessment tasks differ in how much language students must formulate independently, and that difference shapes what an assessment can show about productive language use. This study examines the issue in minority-language assessment by analysing the complete publicly accessible corpus of Grade 8 Slovene INVALSI forms from the first computer-based cycle: two forms comprising 88 task groups. A six-level response-demand framework was developed and applied through qualitative document analysis, descriptive counts, and independent recoding. It distinguished selected and structured responses, local constructed responses, bounded sentences, relational responses, and sustained texts. Thirty-nine task groups (44.3%) required students to formulate Slovene responses. Of these, 23 required a word, phrase, grammatical form, number, or other local unit; 10 required a bounded sentence or formal transformation; and six required a causal, comparative, descriptive, or justificatory relation. No task required students to select, organise, and develop meaning across multiple sentences. The 2019 Italian INVALSI exemplar showed a similar upper boundary, whereas the 2019 Slovenian Grade 9 NPZ required four multi-sentence texts and assessed content and language quality separately. The Slovene forms sample productive language use at local, sentence, and relational levels, but not sustained academic writing. This evidential boundary matters in Slovene-medium education, where Slovene is both a minority language and the language of education. The analysis does not show that students cannot write extended texts or that unequal language exposure affected performance. Evidence of extended writing should therefore be drawn from complementary sources, including the lower-secondary State Examination and classroom writing across subjects.