yao-geo-baidu-keywords
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONOBFUSCATION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script
scripts/build_workbook.pyto process collected keyword data and generate an Excel report. This is a standard functional requirement of the skill and operates on data generated during the session. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data scraped from Baidu (autocomplete suggestions and related searches). It effectively mitigates potential injection risks through explicit instructions to treat page content only as data and through technical controls in the Excel generation script
scripts/build_workbook.py, which forces all scraped text to be treated as literal strings (data_type = 's'), preventing Excel formula injection attacks. - [OBFUSCATION]: A Base64-encoded string is found in the test script
scripts/check_workbook.py. Technical analysis confirms this is a benign 1x1 transparent PNG pixel used as a placeholder to verify workbook generation logic during regression testing. - [SAFE]: The skill implements strong security practices, including domain validation for URLs, path traversal protection for file references, and strict schema validation for its internal data contracts. All external resources and naming conventions are consistent with the vendor's stated identity.
Audit Metadata