correlation-auditor

Pass

Audited by Gen Agent Trust Hub on Nov 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted tabular data files (CSV, TSV, Excel, JSON) where column headers are directly incorporated into natural language summaries in the tool's JSON output.
  • Ingestion points: The scripts/correlation_explorer.py script ingests untrusted data from user-specified file paths.
  • Boundary markers: No delimiters or sanitization steps are used to isolate potentially malicious natural language instructions embedded in column names from the generated interpretation text.
  • Capability inventory: The skill possesses file-writing capabilities via the --output parameter and processes external data files.
  • Sanitization: The script does not escape or validate column names before including them in the 'interpretation' field of the output JSON.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 22, 2026, 01:12 AM