SSH private key accidentally committed in AI agent instruction file

[kasparovabi]

Hi,

I am a security researcher conducting a study on accidental secret exposure in AI agent instruction files (CLAUDE.md, AGENTS.md).

While scanning publicly accessible repositories, I found that SkyflowFoundry/SkyflowForAWSLambda contains what appears to be a real RSA private key inside CLAUDE.md.

File: https://github.com/SkyflowFoundry/SkyflowForAWSLambda/blob/HEAD/CLAUDE.md

The key block (-----BEGIN RSA PRIVATE KEY-----) appears in the repository's agent instruction file. If this key is still in use, anyone who has seen this public file can use it to authenticate to systems that trust it.

Recommended immediate steps:

1. Revoke or rotate this key immediately on all systems where it is authorized.
2. Add the file to .gitignore to prevent future exposure.
3. Consider using git filter-repo or BFG Repo Cleaner to purge the key from git history.

This report is part of coordinated responsible disclosure. I am not exploiting the key and have not accessed any systems with it.

Best regards

[kasparovabi]

Retraction: this report was a false positive

I am withdrawing this report. My scanner matched the string -----BEGIN ... PRIVATE KEY----- as a header pattern without verifying that an actual key body followed it. In this repository the match is a documentation string, a checklist entry, or an example pattern, not committed key material.

I re-verified by requiring a complete PEM structure (header, at least two lines of base64 body, footer) and this file does not contain one. No private key was ever exposed here.

I apologise for the noise and for any time spent investigating. The recommended actions in my original report do not apply and should be disregarded.

Closing as invalid.