The Agent Skills Directory

DATA_EXFILTRATION (SAFE): The skill utilizes curl for service health checks in scripts/check-endpoints.sh and SKILL.md. These operations are targeted exclusively at localhost (ports 9090, 3000, 9093), which is a whitelisted destination for monitoring tools and poses no exfiltration risk.
COMMAND_EXECUTION (SAFE): The provided shell and Python scripts perform benign diagnostic tasks. No patterns of unsafe command interpolation, shell injection, or unauthorized privilege escalation were detected.
PROMPT_INJECTION (SAFE): Analysis of SKILL.md and referenced markdown files found no evidence of instructions designed to bypass agent safety filters or override system prompts.
EXTERNAL_DOWNLOADS (SAFE): No external dependencies, third-party packages, or remote scripts are downloaded or executed by this skill.

monitoring-skill