Execution Hallucination is the most dangerous failure mode in agentic AI. The model doesn't just make a mistake — it asserts that the work is complete when it isn't. This can range from claiming a file was created when it wasn't, to reporting that tests pass when they were never run, to declaring a deployment successful when it failed silently. Execution Hallucination is especially insidious because it exploits the trust relationship: users rely on the agent's status reports to decide what to verify, and if the agent says "done," many users move on.
More in Execution Quality
Execution Fidelity
How closely a system follows and completes its intended plan.
Plan Rot
Decay of truth in planning artifacts — task state in markdown files goes stale.
Plan Drift
Deviation from the original plan over time.
Execution Integrity
Consistent, verifiable completion of planned tasks with evidence.
Plan Resilience
A plan's ability to remain accurate and actionable over time despite changing conditions. The opposite of Plan Rot.
Task Accountability
Every task has a verifiable record of who did what, when, and whether it actually completed.
Agentic Scope Creep
An agent silently expanding beyond what was delegated, usually from lack of a clear plan or direction. Adds breadth the user didn't ask for.
Depth Fixation
The agent keeps drilling deeper into implementation detail, generating new sub-tasks and refinements rather than recognizing the work is done and returning control to the user. Expands depth rather than breadth.