@6171/csv-tools

Pass

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Risk. The skill reads external CSV data and displays its contents to the agent's context. If the CSV file contains instructions formatted to look like system prompts or commands, the agent might interpret them as authoritative instructions.
  • Ingestion points: scripts/csv_tools.py reads CSV files via csv.DictReader in functions cmd_info, cmd_view, cmd_filter, cmd_stats, and cmd_convert.
  • Boundary markers: None. The script directly outputs row content and stats to the terminal, which is then ingested by the agent without specific delimiters or instructions to ignore embedded content.
  • Capability inventory: The skill includes file system write capabilities (cmd_filter, cmd_merge, cmd_convert) and column processing.
  • Sanitization: None. The script reads and prints strings directly from the CSV file without escaping or sanitizing potential command sequences or prompt-like structures.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 5, 2026, 04:12 AM
Security Audit — agent-trust-hub — @6171/csv-tools