TheLibrarian.io web_fetch Tool Exposed Internal Cloud Environment
Mindgard security researchers discovered that TheLibrarian.io's web_fetch tool could be tricked via encoded or non-obvious URL variations to access internal cloud environments, potentially exposing confidential user data. The vendor reportedly did not respond to disclosure attempts.
What happened
TheLibrarian.io's web_fetch tool is designed to retrieve contents from public URLs. The AI refuses requests to private destinations such as localhost. However, by providing encoded or non-obvious URL variations, the tool does not validate the input and proceeds to point its fetching client inward, thereby exposing internal information.
Reported impact
The Mindgard platform identified the vulnerabilities as high severity, exposing TheLibrarian.io's internal cloud environment to infiltration and attack. The issues could be further developed to expose confidential user data.
Index analysis
Mindgard discovered the vulnerability on October 8, 2025 and disclosed it to the vendor on October 10, 2025. The vendor was reportedly incommunicative throughout Mindgard's attempts to coordinate and mitigate the threats, despite having a detailed vulnerability disclosure policy.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.