delivery-tracking

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill uses npx -y @nomadamas/k-skill@0 to download and execute the vendor's CLI tool for fetching instructions and managing helper files.
  • [COMMAND_EXECUTION]: The workflow executes multiple shell commands, including curl for network requests to logistics providers and python3 for processing extracted data. It also uses subprocess.run within Python scripts to invoke system utilities.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted HTML and JSON data from third-party logistics websites (CJ Logistics and Korea Post).
  • Ingestion points: Fetches tracking data from official carrier endpoints via curl commands defined in instruction.md.
  • Boundary markers: Content is parsed and normalized into a structured JSON schema, but input data is not wrapped in explicit security delimiters before processing.
  • Capability inventory: The skill uses curl, python3, and subprocess.run to interact with external sites and local system tools.
  • Sanitization: Employs regex and HTML unescaping to extract specific data fields, which isolates the relevant tracking information from the broader page content.
  • [DYNAMIC_EXECUTION]: The skill generates and executes Python logic at runtime using shell heredocs and temporary files to handle data parsing and HTML extraction tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:11 AM
Security Audit — agent-trust-hub — delivery-tracking