The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses npx to fetch updated instructions and run a local Python script for parsing. All npx calls target the author's own scoped package (@nomadamas/k-skill), which is consistent with the skill's provenance.
[DATA_EXPOSURE]: The skill interacts with the JobKorea recruiting platform but includes explicit 'Hard boundaries' and 'Hard rules' that prohibit the agent from accessing or storing plaintext credentials, session cookies, or candidate personal information without immediate user approval.
[INDIRECT_PROMPT_INJECTION]: The skill processes external HTML data from JobKorea. It implements mitigation by using structured parsing (BeautifulSoup) and regex-based text cleaning (clean_text) to strip script/style tags and action controls, reducing the risk of malicious instructions embedded in candidate profiles influencing the agent.
[SAFE]: The Python scripts (jobkorea_talent_search.py, etc.) are dedicated to search query construction and HTML parsing. They use standard libraries (urllib, json, argparse) and perform targeted AJAX requests to official JobKorea endpoints without suspicious redirection or exfiltration patterns.

jobkorea-talent-search