What problem does it solve?
Transient failures in HTTP calls, RPC, and database operations require consistent retry behavior, and this Skill shows how to define, apply, and manage semantic retry policies for Rust Golem agents without redeploying code.
Core Features & Use Cases
- Manifest-Based Policies: Define named retry policies with priorities, predicates, and composable strategies (exponential, fibonacci, jitter, countBox, timeBox) in golem.yaml per environment.
- Runtime SDK Control: Build and apply policies from agent code using golem_rust::retry, including scoped usage with with_named_policy and querying active policies.
- Live CLI Management: Create, update, list, and delete retry policies at runtime with the golem retry-policy commands, taking effect immediately.
- Use Case: An agent making outgoing HTTPS requests can apply a policy that retries only transient 5xx responses with exponential backoff clamped to [100ms, 5s] and 15% jitter, while a catch-all policy handles everything else.
Quick Start
Ask the AI to configure a retry policy in golem.yaml that retries transient HTTPS errors up to 5 times with exponential backoff and jitter for a Rust Golem agent.