subway-lost-property

Pass

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and update the @nomadamas/k-skill package from the npm registry.
  • [REMOTE_CODE_EXECUTION]: Code from the external @nomadamas/k-skill package is executed to retrieve primary instructions and provide helper functionality.
  • [COMMAND_EXECUTION]: The helper script scripts/subway_lost_property.py executes the curl command via subprocess.run to interact with official lost property search portals.
  • [DYNAMIC_EXECUTION]: The skill uses subprocess.run with list-based arguments to verify the reachability of external websites at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process results from external government sites (LOST112, Seoul Metro). While these are official sources, ingestion of external HTML represents a potential indirect prompt injection surface.
  • Ingestion points: User-provided keywords are used to fetch content from lost112.go.kr in scripts/subway_lost_property.py.
  • Boundary markers: Explicit safety rules in SKILL.md instruct the agent not to take actions without user approval.
  • Capability inventory: Uses subprocess.run to call curl for network access.
  • Sanitization: Employs shlex.quote for command formatting in printed examples and list-based argument passing for execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 1, 2026, 02:41 PM