The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted data through its text and CSV processing functions, which is a common surface for indirect prompt injection if the agent later processes the anonymized output as instructions.
Ingestion points: anonymize() and anonymize_csv() in SKILL.md.
Boundary markers: Absent; the documentation does not specify delimiters to separate data from instructions.
Capability inventory: File system read/write operations for CSV processing.
Sanitization: The tool itself is a PII sanitizer, though it does not explicitly handle instruction-based sanitization.
[EXTERNAL_DOWNLOADS] (SAFE): The dependencies pandas and faker are well-known, trusted libraries from standard package registries.

data-anonymizer