[Feature Request] Anthropic API safety filter incorrectly flags routine system administration commands

Resolved 💬 1 comment Opened Jun 10, 2026 by TimGFF Closed Jun 10, 2026

Bug Description
Fable 5 safety filter incorrectly triggers on routine system administration.
I'm running a read-only architecture audit of my OWN two servers (personal +
office use). Standard diagnostic commands — whoami, id, sudo -n true check,
uname, docker ps, ss -tulpn, ssh connectivity test — get flagged as
cybersecurity content and force a switch from Fable 5 to Opus 4.8 on almost
every command. These are harmless read-only ops on my own infrastructure.
Please refine the classifier so legitimate self-owned sysadmin/audit work
doesn't keep getting flagged.

Environment Info

  • Platform: linux
  • Terminal: vscode
  • Version: 2.1.152
  • Feedback ID: 40733726-e85e-48a8-8939-f92e329b5cac

Errors

[{"error":"Error: Unable to render message type: fallback\n    at KT_ (/$bunfs/root/src/entrypoints/cli.js:4018:37772)\n    at Bf (/$bunfs/root/src/entrypoints/cli.js:551:21158)\n    at Yj (/$bunfs/root/src/entrypoints/cli.js:551:39928)\n    at HKH (/$bunfs/root/src/entrypoints/cli.js:551:50740)\n    at q9 (/$bunfs/root/src/entrypoints/cli.js:551:87775)\n    at E$ (/$bunfs/root/src/entrypoints/cli.js:551:86741)\n    at BWH (/$bunfs/root/src/entrypoints/cli.js:551:86562)\n    at IJ (/$bunfs/root/src/entrypoints/cli.js:551:82895)\n    at oH (/$bunfs/root/src/entrypoints/cli.js:551:6564)\n    at dH (/$bunfs/root/src/entrypoints/cli.js:551:5018)","timestamp":"2026-06-10T10:04:01.818Z"}]

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗