Automatically generating issue-reproducing tests is essential for debugging and regression prevention. Recent advances show that large language models can effectively perform this task, but their reliance on extremely large parameter scales limits practical deployment. In this paper, we propose SLIT, an execution-grounded framework that enables small language models to generate effective issue-reproducing tests. SLIT formulates issue reproduction as an execution-guided exploration problem and optimizes policies with reinforcement learning over execution-based rewards. To improve inference-time exploration, SLIT further introduces diversity-augmented test construction and a learned scoring model to select effective reproducing tests. Experimental results on SWT-Bench-Verified indicate that the 14B-parameter model achieves a resolve rate of 46.2%, achieving competitive performance against prior approaches that rely on larger models. These results demonstrate the effectiveness of SLIT and highlight the viability of execution-grounded small language models for issue-reproducing test generation.

LLM-based agents can interact with external environments through tool invocation, but this capability also introduces security risks such as file modification, information leakage, and unauthorized actions. Existing guardrails often evaluate completed trajectories, leaving pre-execution monitoring of step-level actions underexplored. We propose StepGuard, a step-level guard model that can audit completed agent trajectories and check tool actions before they are executed. To train StepGuard, we introduce StepGen, an automatic data engine that generates safe and unsafe trajectories with the same context but different actions at the risky step. To further reduce over-defense and under-defense, we propose Balance-GRPO, which dynamically balances learning between safe and unsafe actions based on their observed accuracy. Experiments show that StepGuard achieves the highest average accuracy among open-weight guard models, with performance comparable to GPT-5.4. When used to guard agents on AgentDojo and AgentDyn, StepGuard reduces mean ASR by 77.3% relative to the no-guard setting, while mean utility drops by only 2.8 points.