The Agent Skills Directory

[SAFE]: The skill is a defensive tool that facilitates consumer protection by providing structured guidance for identifying scams. It enforces safety best practices, such as explicitly forbidding the agent from confirming a message is safe and directing users to official verification channels. All referenced URLs are to trusted Israeli government and regulatory domains.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to process untrusted data provided by the user (suspected scam messages). Ingestion occurs via the user query. The skill mitigates this risk by providing a strict analytical framework that instructs the agent to treat sender names and message threads as attacker-controlled and to never obey instructions found within the suspected messages. The skill's capability inventory is limited to read-only information retrieval and a local calculation script, posing no threat of exfiltration or privilege escalation.\n- [COMMAND_EXECUTION]: The skill includes a local utility, scripts/liability_clock.py, which is invoked by the agent to compute financial liability figures under Israeli law. The script uses only standard Python libraries (argparse, datetime) and performs purely mathematical operations without file system or network access, ensuring it remains within the secure execution boundaries of the agent platform.

israeli-scam-detector