daangn-realty-search

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions utilize npx to download and run the @nomadamas/k-skill package. This package is managed by the skill's author and serves as a utility for retrieving documentation and executing helper scripts. This behavior is consistent with the vendor's ecosystem for skill management.
  • [COMMAND_EXECUTION]: The skill requires the execution of shell commands through npx to initialize the instruction set and interact with legal disclaimers. These commands are part of the standard operating procedure defined by the vendor.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and processes content from external web pages at realty.daangn.com.
  • Ingestion points: The Python scripts daangn_realty.py and daangn_detail_ld.py fetch HTML and JSON-LD data from Daangn's real estate service endpoints.
  • Boundary markers: The instructions include behavioral rules (e.g., no automatic payments, no credential storage), though there are no explicit delimiters surrounding the ingested web data to separate it from agent instructions.
  • Capability inventory: The skill utilizes network access (urllib.request) and shell command execution (npx) as primary mechanisms for data retrieval and setup.
  • Sanitization: Data parsing is performed using standard json.loads and specific regex patterns to extract structured metadata, which provides a layer of validation against unstructured malicious content injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:12 AM
Security Audit — agent-trust-hub — daangn-realty-search