Survey of 27 Papers Identifies Six Failure Clusters in LLM Agents

A synthesis of 27 benchmark, taxonomy, and audit papers spanning 19 distinct benchmarks (2023-2026) identifies six cross-cutting failure clusters in large language model agents. The taxonomy integrates evidence across tool use, planning, long-horizon reasoning, multi-agent coordination, safety, and measurement validity. The six clusters are: tool invocation and parameter-level errors; planning and constraint-satisfaction failures; long-horizon degradation from context accumulation; multi-agent coordination failures; safety and security failures under adversarial or underspecified conditions; and measurement validity problems. Failures compound nonlinearly with task length, strong performance on individual sub-tasks does not reliably translate to end-to-end success, and additional scaffolding does not consistently improve reliability. Progress has been demonstrated in single-turn tool use, short-horizon web navigation, and narrowly scoped coding tasks. The taxonomy was derived iteratively by grouping independently reported error categories into themes corresponding to distinct stages of the agent reasoning-to-action pipeline.