gov-overseas-trip-report

Pass

Audited by Gen Agent Trust Hub on Aug 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses npx to dynamically download and execute the kordoc and pdfjs-dist packages from the npm registry to facilitate document parsing. These are functional utilities required for the skill's primary purpose of processing PDF and HWP formats.
  • [COMMAND_EXECUTION]: The agent is instructed to execute a localized Python helper script (gov_overseas_trip_report.py) and npm-based utilities to perform data extraction and analysis. These operations are strictly limited to the skill's documented functional scope.
  • [PROMPT_INJECTION]: The skill processes external PDF and HWP documents, which represent a potential indirect prompt injection surface. The risk is mitigated through several factors:
  • Ingestion points: Data is fetched from verified official government websites (.go.kr) and documents are processed locally (SKILL.md).
  • Boundary markers: Explicit instructions are provided to the agent to treat all content from processed documents as data for analysis only, and to ignore any embedded commands or instructions (SKILL.md).
  • Capability inventory: The skill enables specific shell commands for parsing and network requests to official government domains, with no access to sensitive local system credentials (gov_overseas_trip_report.py).
  • Sanitization: The workflow mandates structuring extraction results into specific fields (facts, arithmetic, disclosure, etc.), which helps isolate raw document content from the agent's logic.
  • [SAFE]: The skill exclusively interacts with official South Korean government domains (e.g., nec.go.kr, acrc.go.kr) and includes built-in safety rules for request delays, volume limits, and the automatic deletion of temporary files used during processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 1, 2026, 06:48 AM
Security Audit — agent-trust-hub — gov-overseas-trip-report