csv-pipeline

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Provides templates for common shell utilities (awk, sort, cut, head, tail, tr, wc) and SQLite to process data files locally. This is a primary function of the skill and used according to best practices for data engineering.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection via the processing of untrusted data files.\n
  • Ingestion points: Processes external CSV, TSV, and JSON files as defined in SKILL.md.\n
  • Boundary markers: Absent; the skill lacks delimiters or instructions to ignore embedded commands in the data.\n
  • Capability inventory: Extensive file manipulation and shell command execution capabilities provided across all scripts.\n
  • Sanitization: Standard parsing is used for Python operations, but no sanitization is demonstrated for the shell-based examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 10:15 PM