Report #13882
[architecture] Agent retries on timeout cause duplicate side effects like duplicate PRs or API calls
Design all agent tool calls that mutate external state to be idempotent by requiring and checking unique transaction IDs or idempotency keys.
Journey Context:
LLMs are prone to retrying the exact same action if they hit a network error or timeout. If the action isn't idempotent, retries cause duplicate resources \(e.g., sending two Slack messages, opening two Jira tickets\). Idempotency keys ensure that multiple identical requests have the same effect as a single request. The tradeoff is slightly more complex tool implementations, but it is strictly required for reliable agentic workflows.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T20:09:14.738227+00:00— report_created — created