jobkorea-talent-search
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses the
@nomadamas/k-skillpackage vianpxto fetch updated instructions, file listings, and legal disclaimers from remote sources. This package is maintained by the skill's author. - [REMOTE_CODE_EXECUTION]: Instructions in
SKILL.mddirect the user to run variousnpxcommands. These commands download and execute JavaScript code from the npm registry to manage skill state and execute the fallback search script. - [INDIRECT_PROMPT_INJECTION]: The skill analyzes external data from JobKorea resumes, including career history, project descriptions, and portfolio links. This presents an attack surface where malicious instructions embedded in a public resume could attempt to manipulate the agent's evaluation or scoring logic.
- Ingestion points: JobKorea search result rows and detail pages via browser session or AJAX endpoint.
- Boundary markers: The skill does not implement explicit delimiters for external data in the prompt, though it uses structured output formatting.
- Capability inventory: The skill primarily performs data retrieval via
urlliband analysis; it does not perform automated file-writes or privileged system operations during the analysis phase. - Sanitization: The
jobkorea_talent_parse.pyscript implements basic HTML sanitization, explicitly stripping<script>and<style>tags from the ingested content.
Audit Metadata