claude-code - 💡(How to fix) Fix Claude Code lacks independent reasoning — defaults to docs/assumptions instead of running available tools first [2 comments, 2 participants]

TL;DR

Run the verification command gsd --list-models <provider> before applying config changes based on documentation to ensure accuracy for the specific provider.

Guidance

• Verify the available models for the actual provider (in this case, GitHub Copilot) using the gsd --list-models command to avoid making incorrect assumptions based on documentation.
• Check the documentation for the specific provider being used, as the correct notation may vary (e.g., dots for GitHub Copilot, dashes for Anthropic).
• Before applying any config fixes, confirm that the proposed change aligns with the output of the verification command.
• Consider establishing a practice of running verification commands as a preliminary step when troubleshooting or making config changes.

Example

No code snippet is necessary for this issue, as the focus is on the process of verification rather than code implementation.

Notes

The approach should be tailored to the specific provider and model in use, as different providers may have different notation conventions. This guidance assumes that a verification command like gsd --list-models <provider> is available and applicable.

Recommendation

Apply workaround: Run the verification command before making config changes. This ensures that any fixes are based on the actual models available for the provider in question, rather than potentially outdated or incorrect documentation.