The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions attempting to override agent behavior or bypass safety filters were found in the skill or documentation.
[Data Exposure & Exfiltration] (SAFE): No patterns of accessing sensitive file paths or performing unauthorized network operations were detected.
[Obfuscation] (SAFE): No hidden, encoded, or obfuscated content such as Base64 blocks or zero-width characters was found.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill utilizes well-known, trusted data science libraries (NumPy, SciPy, pandas) and does not perform any remote script downloads or runtime package installations.
[Privilege Escalation] (SAFE): There are no commands that attempt to gain elevated privileges or modify system-wide permissions.
[Persistence Mechanisms] (SAFE): No mechanisms for maintaining persistent access, such as shell profile modifications or scheduled tasks, were identified.
[Metadata Poisoning] (SAFE): The metadata is accurate, transparent, and contains no hidden malicious instructions.
[Indirect Prompt Injection] (SAFE): While the skill is designed to process external numerical data, the risk is mitigated by the lack of write operations, network access, or side-effect capabilities.
[Time-Delayed / Conditional Attacks] (SAFE): No logic-gated or time-delayed malicious triggers were found in the codebase.
[Dynamic Execution] (SAFE): The skill uses safe configuration loading (yaml.safe_load) and avoids dangerous dynamic evaluation functions like eval() or unsafe deserialization.

statistics-math