anysite-competitor-intelligence
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of an
execute()tool pattern. However, this is defined as a 'universal meta-tool' belonging to the vendor's MCP server for data fetching (LinkedIn, YC, Social Media) rather than arbitrary shell execution. The parameters passed are structured arguments for API endpoints. - [DATA_EXFILTRATION]: While the skill involves collecting data from various platforms (LinkedIn, Twitter, Reddit), this is the stated primary purpose of the 'Competitor Intelligence' tool. There is no evidence of harvesting local sensitive files (like SSH keys or AWS credentials) or exfiltrating user-private data to an unauthorized third party.
- [EXTERNAL_DOWNLOADS]: The skill uses
webparserandduckduckgotools to fetch content from the web for analysis. These are legitimate tools for web scraping and search within the context of competitive research and do not involve downloading or executing untrusted code. - [PROMPT_INJECTION]: No evidence of instructions designed to bypass agent safety filters or override system instructions was found. The instructions are focused on guiding the agent through complex multi-step workflows for data gathering.
- [REMOTE_CODE_EXECUTION]: No patterns of remote script execution (e.g.,
curl | bash) or dynamic code loading from untrusted sources were detected.
Audit Metadata