Fable 5 safety banner advises /feedback, but /feedback is unavailable in the VS Code extension (dead-end feedback path)

Summary

In the Claude Code VS Code extension, the Fable 5 safety banner instructs users to "Send feedback with /feedback" — but /feedback is not available in the VS Code extension (it returns "/feedback isn't available in this environment."). Per the docs, /feedback exists in the terminal CLI but not the IDE extension, so the banner gives environment-blind advice: following its own instruction fails.

Environment

• Surface: Claude Code VS Code native extension
• Model: Fable 5 (claude-fable-5[1m]), auto-switched to Opus 4.8 by the safety measure
• OS: macOS

Steps to reproduce

1. Use the Claude Code VS Code extension on the Fable 5 model.
2. Send a message that trips the Fable safety measure so the banner appears (see verbatim text below).
3. Follow the banner's instruction and run /feedback.
4. Actual: /feedback isn't available in this environment.

Expected

The safety banner should offer a feedback path that works in the current surface — e.g., a clickable link/button, or environment-appropriate wording — rather than naming a slash command (/feedback) that doesn't exist in the VS Code extension. As written, the only remediation the banner advertises is a dead end for IDE users.

Banner text (verbatim)

Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://support.claude.com/en/articles/15363606

Secondary note (model behavior — false positive)

The flag fired on a routine, authorized defensive security review of my own infrastructure (a findings/remediation summary). Per the banner's own admission that it "may flag safe, normal content as well," this is a false positive that interrupts legitimate security work with a forced model switch. Requesting tuning so authorized defensive/audit content isn't swept up — but the primary, clearly-reproducible defect is the broken feedback path above.