openai-whisper-api
Fail
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/transcribe.shperforms unsafe shell expansion when invokingcurl. Specifically, variables expanded using the${var:+...}syntax (such as for--languageand--prompt) are not double-quoted. This leads to word splitting, which an attacker can exploit by providing an argument containing spaces to inject additional flags into thecurlcommand. - [DATA_EXFILTRATION]: Due to the argument injection vulnerability in the
curlcommand, an attacker can use the-F "key=@file"syntax to read and exfiltrate sensitive files from the local filesystem (such as.envfiles or SSH keys) by sending them to the API endpoint or an injected malicious destination.
Recommendations
- AI detected serious security threats
Audit Metadata