Tool-use failures
- Wrong or missing tool call
- Incorrect or incomplete arguments
- Schema ambiguity
- Response misunderstood
- Permission or authentication failure
- Duplicate or conflicting actions
RESOURCE / FAILURE TAXONOMY
Find the class of failure behind the symptom.
A working taxonomy for common, expensive, and diagnosable failures in production tool-using agents.
EARLY ACCESS SCOPE
A production AI agent failure taxonomy.
Early access is focused on failures that are common, expensive, and diagnosable.
Context and memory failures
- Relevant context omitted
- Irrelevant context dominated
- Wrong retrieval evidence
- Stale memory overrode current state
- Conflicting facts unresolved
- Important history lost
Planning and orchestration failures
- Wrong skill or sub-agent
- Plan failed to adapt
- Failed step not recovered
- Retries caused loops
- Timeout or budget termination
- Incorrect stop criteria
Evaluation failures
- Style scored over correctness
- Judge contradicted business outcome
- Deterministic condition missing
- Neighboring regression untested
- Dataset drifted from production
HOW TO USE IT
Classify the earliest unrecovered failure.
Start from the observed business outcome, reconstruct the execution path, and identify the first point where behavior departed from the expected state and was not corrected downstream.
Keep observed facts separate from likely explanations, alternative hypotheses, missing evidence, and confidence. Convert each meaningful incident into a reusable regression case.
START WITH EVIDENCE
Bring us one agent failure your team could not explain quickly.
No generic sales demo. Start with a real production incident from your agent.