analyst-estimates
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (MEDIUM): The skill requires the installation of the
octagon-mcppackage vianpx. This source is not on the trusted list of organizations, making it an unverifiable dependency. - REMOTE_CODE_EXECUTION (MEDIUM): The
octagon-mcpserver is executed dynamically usingnpx -y, which fetches and runs code at runtime, presenting a risk if the package is compromised. - COMMAND_EXECUTION (LOW): Installation instructions in
references/mcp-setup.mdinclude acurl | bashcommand for Homebrew. While standard for that tool, it is an insecure pattern for remote script execution. - INDIRECT_PROMPT_INJECTION (LOW): The skill interpolates user-provided tickers directly into tool prompts without sanitization. Evidence: 1. Ingestion:
<TICKER>and<N>variables inSKILL.md. 2. Boundaries: Absent in the JSON MCP payload. 3. Capability: Access to theoctagon-agenttool. 4. Sanitization: None defined in the skill instructions.
Audit Metadata