Rank, 14 days: 62 on the Claude Code sheet
3 people asked for

All time, since Nov 2, including 1 thumbs-up on GitHub.

Stop cyber classifier downgrading models during authorized security testing

“Repeated false positives on authorized pentesting work (forced model downgrade)”

What gets counted gets built. Pile on.

Paste this into your agent. It files your request, with what you need it for.

I want this: rapsheet.dev/sheets/claude-code/issues/24196/file
Follow it to file my request. Show me the text before you send.

No install. No account. Public and anonymous. The link asks your agent for one report, nothing else. What it sends

Reporters doing authorized pentesting and white-box assessments of their own apps are switched from Opus 5.5 to earlier models, even on harmless messages.

Is this your request?

File my request
Where
Cyber safety classifier
Versions
2.1.316
Platforms
Linux, macOS

Related: Work past cyber safeguard blocks

Not quite your request?

  1. Cyber flags on development prompts8 people
  2. Maintain owned or client servers without cyber safeguard blocks2 people

Original reports2, newest first

  1. Reduce over-restrictive safeguards blocking pentesting of user-owned web apps#107061
  2. Allow declared authorized-engagement context to prevent cyber classifier false-positive model downgrades#106766

Do you work on Claude Code?

Claim its RAP Sheet to see what your users report and answer them in one place.