korean-patent-search
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions in
SKILL.mdandinstruction.mdrecommend usingnpx -y @nomadamas/k-skill@0to fetch and execute updated instructions and helper scripts from the NPM registry. - [COMMAND_EXECUTION]: The skill requires the execution of shell commands, specifically
npxandpython3, to run the patent search logic provided inscripts/patent_search.py. - [DATA_EXFILTRATION]: The helper script
scripts/patent_search.pytransmits theKIPRIS_PLUS_API_KEY(sourced from environment variables or command-line arguments) to the official KIPRIS Plus API endpoint atplus.kipris.or.krto authenticate requests. This is a functional requirement for the service. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source (patent abstracts and titles from the KIPRIS API).
- Ingestion points: External XML data is fetched via
urllib.request.urlopeninscripts/patent_search.py. - Boundary markers: The script does not implement specific boundary markers or sanitization for natural language instructions that might be embedded in patent text.
- Capability inventory: The script's capabilities are limited to network reads and printing JSON to standard output; it does not perform file writes or system command execution based on the API response.
- Sanitization: Basic whitespace cleaning is performed via
clean_text, but no content filtering for prompt injection is implemented.
Audit Metadata