The failure taxonomy
Agents do not crash. They produce confident, well-formatted, wrong answers, and the failure travels downstream undetected.
Agents are probabilistic systems doing deterministic work. When they fail they do not crash; they produce confident, well-formatted answers while every metric reads green. Three modes feed one amplifier: tool misuse, step repetition and reasoning-action mismatch, all flowing into error propagation.
The same taxonomy at working resolution: what each mode looks like from the outside, and which harness mechanism catches it. The amplifier is the one to fear; the modes are merely where it starts.
Probabilistic systems, deterministic work
Traditional software is deterministic, which implies same input, same output, and when it does fail it gives off concrete signals. An exception, a log entry, an alert. The bug is found, fixed and a patch is deployed. Thus the failure does not happen again. Agents break this paradigm. They are probabilistic systems doing deterministic work. The same input may succeed nine out of ten times and fail on the tenth. And when agents do fail, they do not crash. They produce confident, well-formatted and authoritative sounding answers. The system appears healthy and every metric reads green. The failure travels downstream undetected.
The modes
In literature there are already attempts to classify types of failures when it comes to agents. The most significant are the following.
Tool misuse happens when the agent calls the right tool with the wrong arguments. This can occur because the agent "feels" it is correct to invent a parameter or entity that does not exist. The operation completes with no results and the agent reports this as a fact, with no errors raised. The user blames data quality.
When an agent repeats the same action without progressing it gets stuck in a loop. This is called step repetition and it causes increased compute consumption while to the telemetry the agent appears to be working. The reality is that the agent is running in circles.
Reasoning-action mismatch occurs when the agent reasons one thing and does another. Its internal chain-of-thought reasons to do one thing and yet it proceeds to execute another. The reasoning can be sound but the execution did not follow it. According to the MAST multi-agent failure taxonomy this represents roughly 14% of agent failures.
The amplifier
Any single failure mode is important and can cause serious issues. However, error propagation is far worse, since it produces false artifacts of success in all downstream steps after the system has already failed. Traditional observability cannot cope with those scenarios because agent failures are semantic and not technical. They look like valid data, not error codes. As such monitoring is blind to them because the system is working. Its answers are wrong.
The tools that detect and prevent agent failures today: evaluation platforms, hallucination detectors, red-teaming frameworks, and vulnerability scanners.
Dominant now
Enterprise AI evaluation and observability platform combining cost tracking, prompt experimentation, and eval-based quality checks.
evaluation platformsEvaluation intelligence platform with Luna SLMs for scalable evals, production guardrails, and agent reliability monitoring.
hallucination detectionOpen-source Python Risk Identification Toolkit for red teaming generative AI with adaptive multi-turn attack loops.
red teamingNew arrivals
Automated evaluation platform building digital worlds that stress-test AI agents with adversarial red-teaming.
adversarial agent testingOpen source picks
Open-source LLM observability and evaluation with distributed tracing via OpenTelemetry and LLM-as-judge evals.
evaluation platformsOpen-source LLM vulnerability scanner probing for hallucination, data leakage, prompt injection, and jailbreaks across 37+ modules.
vulnerability scanningOpen-source AI engineering platform for LLM tracing, evaluation, prompt management, and cost attribution.
trace-based failure analysisOpen-source CLI for LLM evaluation, comparison, and red teaming with 50+ vulnerability types and CI/CD integration.
red teaming| Claim | Source | Status |
|---|---|---|
| Error propagation produces false artifacts of success in all downstream steps after the system has already failed; agent failures are semantic, not technical. | Why AI Agents Break: Production Failures | verified 2026-07-02 |
| Reasoning-action mismatch represents roughly 14% (13.98%) of agent failures in the MAST multi-agent failure taxonomy. | Why Do Multi-Agent LLM Systems Fail? (MAST) | verified 2026-07-02 |