I think this is a Claude issue and being stupid, but it seems to always happen now. It attempts to use strings like "1" instead of integer value 1 - And then the command fails, and annoyingly Claude doesn't seem to retry it just moves on without using sequential-thinking.
Here's a snippet:
next_thought_needed: "true", thought_number: "1", total_thoughts: "10",
is_revision: "false")
⎿ Error: {
"error": "Invalid thoughtNumber: must be a number",
"status": "failed"
}
Even if it is a Claude issue, would there be a way to make the sequential-thinking model more flexible, so it tries to parse as an int if it's in a string format or something?