esg-benchmark-comparison
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMNO_CODEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- NO_CODE (SAFE): The skill files consist entirely of Markdown and JSON metadata. No scripts, binaries, or executable files are included in the package.
- EXTERNAL_DOWNLOADS (MEDIUM): The skill relies on the 'octagon-mcp' server, which is installed via npx from the npm registry. The organization 'OctagonAI' is not recognized as a trusted source in the defined security scope.
- REMOTE_CODE_EXECUTION (MEDIUM): Setup documentation for macOS includes a piped bash command (curl | bash) to install Homebrew. This pattern is flagged as high-risk for remote code execution, although it is a standard practice for this specific tool.
- PROMPT_INJECTION (LOW): The skill exhibits an indirect prompt injection surface (Category 8). 1. Ingestion points: Data returned from the 'octagon-agent' and 'octagon-web-search-agent'. 2. Boundary markers: Absent. 3. Capability inventory: Access to web search and deep research tools. 4. Sanitization: None described.
Audit Metadata