detecting-data-anomalies
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process external datasets, which represents a potential attack surface for indirect prompt injection if those datasets contain malicious instructions.
- Ingestion points: Data ingestion occurs via the
Readtool, loading CSV, JSON, and Parquet files as described inSKILL.mdandreferences/implementation.md. - Boundary markers: There are no explicit instructions or delimiters in the skill body to warn the agent to ignore any potential natural language instructions embedded within the processed data records.
- Capability inventory: The skill leverages the
Bash(python:*)tool to run bundled scripts (data_loader.py,algorithm_selector.py,anomaly_visualizer.py) that iterate through the untrusted data. - Sanitization: While the skill provides comprehensive steps for statistical cleaning (missing value imputation and feature scaling), it lacks specific security sanitization to prevent data content from being interpreted as agent commands.
Audit Metadata