Synthesis of LLM Agent Failures: A Unified Taxonomy of Tool-Use, Planning, and Reasoning Limitations

This paper synthesizes 27 benchmark, taxonomy, and audit papers (2023-2026) spanning 19 distinct benchmarks into a unified taxonomy of LLM agent limitations. It identifies six failure clusters: (1) tool invocation and parameter-level errors, (2) planning and constraint-satisfaction failures, (3) long-horizon degradation from context accumulation, (4) multi-agent coordination failures, (5) safety and security failures, and (6) measurement validity problems. Key findings include that failures compound nonlinearly with task length, strong sub-task performance doesn't guarantee end-to-end success,