trial-listing-draft
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses local Python scripts (
check_listing.pyand_common.py) that rely exclusively on the Python standard library. No external package installations or network-based execution patterns are present.\n- [COMMAND_EXECUTION]: The skill invokes a local Python script to perform listing checks. This is a legitimate functional use of a deterministic tool and does not involve elevated privileges or dangerous shell commands.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted job listing text, presenting a potential indirect prompt injection surface. This is mitigated by the design of the included Python scripts, which filter the input and return only specific matched substrings to the agent context.\n - Ingestion points: The
listing.textfield in the input JSON passed toscripts/check_listing.py.\n - Boundary markers: The skill provides a clear report format in
references/report-format.mdand emphasizes summarizing findings rather than echoing raw input.\n - Capability inventory: Execution of bundled Python scripts and reading of user-provided listing files.\n
- Sanitization: The
scan_textfunction incheck_listing.pyuses regular expressions to extract only specific problematic phrases, preventing the full untrusted content from being echoed to the agent output buffer.\n- [SAFE]: External references inreferences/source-checks.mdpoint to official Japanese government and service-specific domains for regulatory information, which is consistent with the skill's primary purpose.
Audit Metadata