Permission to Retry: A Study of Authorization and Recovery in AI Agent Workflows
Source
arxiv.org
Author
Deekshith Marla, Kushagra Bhatnagar
Date
Why it matters
Shows that retrying a consequential AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → action after authority changes is unsafe unless permission is revalidated and an effect ledger is kept. Naive cached retry passed only 30 of 42 histories.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
articleAgentR A Stateful and Recovery-Aware Software Architecture for LLM-based Auditable WorkflowsRiya Samanta, Bidyut Saha, Soumya Kanti Ghosh, Rajkumar Buyya
articleFinishing the Task Is Not Enough: Evaluating Agent Resilience and Considerate Participation under Accumulating ChallengeYuanchen Bai, Zijian Ding, Angelique Taylor
articleCan MCP Clients Decide What to Do After Failure? A Result-Only Actionability AuditRishabh Mehan