The Agent Skills Directory

[SAFE]: The skill performs legitimate file monitoring and entropy analysis for ransomware detection. The operations are local and consistent with the stated purpose. The instructions and scripts focus on defensive monitoring and behavioral detection logic.
[DATA_EXFILTRATION]: No network activity was detected. The script reads file content for analysis purposes and saves results locally to a JSON file. It does not attempt to send data to external servers or access sensitive configuration files like SSH keys or environment secrets.
[INDIRECT_PROMPT_INJECTION]: The script reads local files to calculate entropy and hashes, which represents a data ingestion surface.
Ingestion points: Files are read in scripts/agent.py via analyze_file_entropy and snapshot_directory_state.
Boundary markers: Absent.
Capability inventory: Local file read and write access; no network or subprocess execution capabilities exist in the provided script.
Sanitization: None, as data is processed as raw bytes for mathematical and cryptographic analysis.

detecting-ransomware-encryption-behavior