What Problem This Solves

In src/link-understanding/runner.ts, the timeout for the guarded HTTP fetch operation (fetchLinkContent) is resolved from the first entry of the link models list: const timeoutMs = resolveTimeoutMsFromConfig({ config, entry: entries[0] });. If there are multiple entries with different execution timeout values, the fetch operation uses the first one. Fetching content is independent of the CLI execution and should use a dedicated global fetch timeout configuration or default timeout rather than capping it dynamically on the first command entry.

Expected vs Actual Behavior

• Expected: The HTTP fetch timeout should be resolved using a global timeout setting or the maximum of the configured entry timeouts, rather than hardcoding the check to entries[0].
• Actual: Fetch requests can fail prematurely on later model entries if they have larger timeouts than the first entry.

Code Link

src/link-understanding/runner.ts#L223-L233

Proposed Solution

Resolve the fetch timeout using a fallback hierarchy starting from config.timeoutSeconds before falling back to the individual model entry timeouts.