The Agent Skills Directory

[SAFE]: The skill instructions and metadata align with its stated purpose of threat hunting. No prompt injection or behavior override patterns were found.
[DATA_EXPOSURE]: The scripts read local network logs (Zeek, CSV, JSON) and generate local reports. No sensitive credential access or unauthorized network exfiltration was identified.
[OBFUSCATION]: No obfuscated strings, multi-layer encoding, or homoglyph attacks were detected. The domain allowlists in the scripts consist of well-known and legitimate services.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute code from remote sources. It references well-known security tools like RITA and Zeek, which are industry standards for this use case.
[COMMAND_EXECUTION]: The provided Python scripts use standard libraries and do not spawn shell processes or execute arbitrary commands.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external log data, it does so using strictly defined statistical models and standard type conversion (float/int). The impact is limited to local report generation, and no interpolation into LLM prompts is performed within the scripts.

hunting-for-beaconing-with-frequency-analysis