Bug Description

When agents.defaults.thinkingDefault is set to "xhigh" and the default model is not one of the xhigh-supported models (e.g. anthropic/claude-opus-4-5), cron jobs using the isolated-agent path crash with:

Error: Thinking level "xhigh" is only supported for openai/gpt-5.2, openai-codex/gpt-5.2-codex or openai-codex/gpt-5.1-codex.

Root Cause

The interactive chat path (src/auto-reply/reply/get-reply-run.ts) gracefully downgrades xhigh → high when the model does not support it. However, the cron isolated-agent path (src/cron/isolated-agent/run.ts) throws a hard error instead.

Relevant code:

• Interactive (graceful): get-reply-run.ts — checks supportsXHighThinking(), falls back to "high"
• Cron (crashes): run.ts — checks supportsXHighThinking(), throws Error

Expected Behavior

Cron jobs should gracefully downgrade xhigh to high (matching the interactive path behavior) instead of crashing.

Steps to Reproduce

1. Set agents.defaults.thinkingDefault to "xhigh" in openclaw.json
2. Set default model to a non-GPT-5.2 model (e.g. anthropic/claude-opus-4-5)
3. Create a cron job without an explicit thinking override
4. Wait for the cron job to fire
5. Job fails with the error above

Workaround

Explicitly set thinking on individual cron jobs, or change thinkingDefault to "high".