Ceiling Effects and Convergence: Null Results for Instruction Repetition in LLM-Agent Pipelines

Prompt repetition, the verbatim duplication of an input transforming <QUERY> into <QUERY><QUERY>, has been shown to improve accuracy for non-reasoning large language models on retrieval and multiple-choice benchmarks (Leviathan et al., 2025). We ask whether applying the analogous <QUERY><QUERY> pattern to fixed delegate instructions produces similar gains in multi-step agentic pipelines. We present three pre-registered controlled experiments in which Claude Haiku 4.5 delegates (n=5 per condition, temperature 0.5) were assigned either a single-copy or a repeated-prompt instruction under blinded binary rubric scoring, totalling 30 sessions and 4,306 messages. Experiment 1 (session-ID refactoring, 6 criteria) yielded a non-significant score delta of +0.30 with five of six criteria saturated at 100% in both groups (Fisher's exact p=1.000). Experiment 2 (tree-sitter scanner evaluation, 7 criteria) produced a complete ceiling effect with all 10 runs scoring 7/7 (Mann-Whitney U=12.5, p=1.000). Experiment 3 (Kotlin grammar synthesis, 7 criteria) revealed a distinct failure mode: three of seven criteria scored 0/1 across both groups due to rubric-runner co-design failure, where criteria required investigations not present in the runner prompt. On the four reachable criteria, control scored a mean of 2.60/4 and treatment scored 2.80/4 (U=15, p=0.607). Across all three pilot experiments, we detect no effect of prompt repetition on task success. Treatment agents used 30.6% fewer tokens in Exp1 but 7.2% and 21.3% more in Exp2–3; the direction reverses with session length, as long sessions save turns while short or saturated sessions pay pure prompt overhead. Criteria requiring investigations absent from the runner prompt are unreachable by both groups and must be resolved before effect sizes are interpretable. We formally characterize rubric-runner co-design failure as a distinct evaluation failure mode in which rubric criteria are structurally unreachable from the runner prompt, and propose a rubric-runner alignment score as a necessary pre-registration gate. This is a preprint; it has not undergone peer review.