The Agent Skills Directory

[Prompt Injection] (SAFE): The content consists entirely of technical documentation and Python code snippets for data engineering. There are no attempts to override system instructions or bypass safety filters.
[Data Exposure & Exfiltration] (SAFE): The skill discusses accessing sensitive file paths like ~/.aws/credentials only in the context of explaining how Pandas auto-detects credentials. No code snippets attempt to exfiltrate these files or hardcode credentials.
[Remote Code Execution] (SAFE): No remote script downloads or piped execution patterns (e.g., curl|bash) were detected.
[Indirect Prompt Injection] (LOW):
Ingestion points: Data entering through pd.read_csv, pd.read_parquet, etc., from cloud URIs.
Boundary markers: Not explicitly defined in snippets.
Capability inventory: Data processing and local file I/O.
Sanitization: Not present in the snippets.
Assessment: The skill creates an attack surface by reading untrusted external data, which is standard for data engineering tasks. The documentation itself is benign.

data-engineering-storage-remote-access-integrations-pandas