access-review-report
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a helper script
make_pdf.pythat executes a local web browser (such as Chrome, Chromium, or Edge) viasubprocess.runto convert generated HTML reports into PDF format. This execution is limited to specific browser binaries found on the system path or specified locations and uses secure argument passing that avoids shell invocation risks. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data retrieved from the
polar-sky-controlsMCP server. To mitigate potential injection risks from this untrusted data, the skill employs several layers of defense: - It instructs the AI to treat all names, reasons, and notes within the tool results strictly as data and never as instructions.
- The
render_report.pyscript programmatically escapes HTML entities in the report output to prevent Cross-Site Scripting (XSS). - The script implements CSV injection defenses (following OWASP guidance) by prefixing potentially dangerous characters (like
=,+,-,@) with a single quote to prevent spreadsheet software from executing them as formulas.
Audit Metadata