oathe / scan / reports / dangerous
Is ankylala/document-parser safe?
No. ankylala/document-parser is dangerous to install. Oathe's behavioral security audit gave the document-parser skill by ankylala a trust score of 63/100 with 5 findings (2 critical).
https://clawhub.ai/ankylala/document-parser
Is ankylala/document-parser safe to install?
This skill presents a critical data exfiltration risk by sending all user documents to an external server without proper disclosure or consent. While it appears to be a legitimate document parsing tool, it actually functions as a data collection mechanism that could be used for corporate espionage or theft of sensitive information.
What security issues were found in ankylala/document-parser?
Category Scores
Findings (5)
CRITICAL Documents sent to external server without consent -100 ▶
The skill sends all user documents to an external server at http://47.111.146.164:8088/taidp/v1/idp/general_parse for processing. This creates a serious data exfiltration risk as sensitive documents could be logged, stored, or misused by the third-party service.
Evidence: index.py line 18: DEFAULT_BASE_URL = "http://47.111.146.164:8088" and parse_document() function uploads files via requests.post()
CRITICAL Deceptive functionality disguised as local tool -100 ▶
The skill presents itself as a document parser but actually functions as a data exfiltration tool. Users would reasonably expect local document processing but instead their files are sent to unknown external servers.
Evidence: Documentation describes document parsing features but implementation uploads files to 47.111.146.164:8088
MEDIUM External network connections during installation -20 ▶
Skill made connections to external endpoints during installation phase, consistent with documented HTTP request functionality.
Evidence: Network monitoring shows connections to 216.150.1.1:443 and DNS queries
LOW HTTP requests to hardcoded external endpoint -15 ▶
Code makes HTTP requests to hardcoded external server, creating potential for abuse if endpoint is compromised.
Evidence: requests.post() calls to http://47.111.146.164:8088/taidp/v1/idp/general_parse
LOW Documentation in non-English language -5 ▶
Primary documentation is in Chinese which could potentially be used to obscure malicious instructions, though content appears legitimate.
Evidence: SKILL.md contains Chinese text describing document parsing functionality
Should I install ankylala/document-parser?
Oathe's verdict for ankylala/document-parser is DANGEROUS with a trust score of 63/100. Recommendation: Do not install.