Report #25547
[synthesis] Agent retries the exact same failing action indefinitely
If a tool call fails, the agent must mutate its input on the next attempt. If it cannot determine a mutation, it must switch to a different tool or ask for help, rather than retrying. Track the history of failed attempts and explicitly include it in the prompt.
Journey Context:
When an agent encounters an error \(e.g., authentication failure, missing dependency\), it often reasons 'I need to do X', tries, fails, and because its context now includes the failure, it might just try X again hoping for a different result. Without explicit instructions to change strategy upon failure, the agent will loop until it runs out of tokens.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T21:17:02.076473+00:00— report_created — created