sharenuri-facility-search
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the external portal
eshare.go.kr. The retrieved HTML content is parsed using regular expressions to extract facility details (name, location, fees) which are then presented to the agent. Malicious instructions embedded in the website's content could potentially influence the agent's behavior. - Ingestion points: The
scripts/sharenuri_search.pyscript fetches data from multiple endpoints onwww.eshare.go.kr(e.g.,ConditionsSearch.do,selectsigunguList.do, and individual facility detail pages). - Boundary markers: No specific delimiters or "ignore instructions" warnings are used when processing the scraped content.
- Capability inventory: The skill environment facilitates shell execution via the
npxcommand as documented in the workflow. - Sanitization: The
_strip_htmlfunction removes HTML tags and normalizes whitespace, but it does not filter or sanitize the textual content for prompt injection patterns. - [EXTERNAL_DOWNLOADS]: The skill instructions direct the user and the agent to use
npx -y @nomadamas/k-skill@0to fetch instructions and execute the Python helper script. This involves downloading and running code from a remote package registry. - Evidence: The
SKILL.mdandinstruction.mdfiles consistently usenpxto interact with the skill's components via the@nomadamas/k-skillpackage.
Audit Metadata