excel

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • PROMPT_INJECTION (HIGH): The skill presents a significant indirect prompt injection surface due to the combination of untrusted data ingestion and file-modification capabilities. 1. Ingestion points: Content is read from external files via read, cell, from-csv, and from-json commands. 2. Boundary markers: The documentation identifies no delimiters or markers to protect against instructions embedded in data. 3. Capability inventory: The skill can modify the file system using commands like write, edit, and delete-sheet. 4. Sanitization: No sanitization or validation of external content is documented.
  • COMMAND_EXECUTION (LOW): The skill operates by executing a local Python script (scripts/excel.py) using the python3 command. This is a standard but noteworthy pattern involving host-level execution.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:51 PM