Increase Daily Rig Decomposition Benchmark timeout

[Copilot]

The daily decomposition benchmark exceeded its 45-minute agent execution limit. This updates the workflow to allow the requested 55-minute budget.

• Workflow timeout
  • Set timeout-minutes: 55 in the benchmark workflow source.
• Generated workflow
  • Regenerated the lock file so agent and failure-report timeout settings remain aligned.

timeout-minutes: 55

• Fixes [aw] Daily Rig Decomposition Benchmark timed out #391

[pelikhan]

@copilot review the agent logs and investing why the agentic workflow timed out

[Copilot]

@copilot review the agent logs and investing why the agentic workflow timed out

The 45-minute limit was consumed by the workflow agent debugging its own benchmark harness, not by one benchmark execution. The first typecheck failed at 02:54:14 because the installed skill lacked Node type declarations. It then spent ~28 minutes investigating SDK authentication/model transport and locally patching the installed rig.ts; subsequent benchmark attempts failed with the SDK’s default 60-second session.idle timeout and then empty/invalid coordinator JSON. It started another diagnostic run at 03:33:44, and the outer workflow terminated it at 03:37:44. The run used 172,155 output tokens (653.713 AIC). Raising the outer limit alone does not address this root cause; the benchmark needs bounded setup/debugging and a harness path that succeeds without local runtime patching.

[pelikhan]

@copilot review the agent logs and fix rig.ts so that sdk issues are not a problem

[Copilot]

@copilot review the agent logs and fix rig.ts so that sdk issues are not a problem

Fixed in 480eb54. Rig now forwards the workflow SDK connection token and provider configuration, honors the configured send timeout instead of the SDK’s 60-second default, and aborts in-flight SDK requests on cancellation.