Allow users to configure the per-tool execution timeout, which is currently hardcoded to 300 seconds. This timeout is frequently hit during complex autonomous workflows, leading to wasted time and context.
### Use case ## Problem The CLI executor in `ralph-adapters` has a hardcoded 300-second (5 minute) timeout per tool execution (`timeout_secs=300`). When running complex autonomous workflows with large codebases, this timeout is frequently hit — especially during: - Large file reads/writes - Running `test` on big projects - Complex LLM responses that take longer to stream Ralph recovers by moving to the next iteration, but the lost iteration wastes time and context. ### Proposed solution ## Proposal Make the per-execution timeout configurable in `ralph.yml` under `cli` or `event_loop`: ``` cli: backend: "kiro" execution_timeout_secs: 900 # 15 minutes ``` Or as a top-level option: ``` event_loop: execution_timeout_secs: 900 ``` Default can stay at 300s for backward compatibility. ### Area CLI ### Backend(s) impacted Kiro ### Alternatives considered _No response_ ### Additional context _No response_