docx-perfect

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): Vulnerable to Indirect Prompt Injection. The skill ingests untrusted content from external Word documents and uses it to generate logic and data for table creation. * Ingestion points: SKILL.md and scripts/template.py read paragraphs from external .docx files using Document('source.docx'). * Boundary markers: Absent. No instructions are provided to the agent to ignore instructions embedded in the document text. * Capability inventory: Shell command execution (python -c), file modification/creation (doc.save), and script execution. * Sanitization: Absent. Text extracted from documents is used directly to define TABLE_DATA or identify sections for replacement in script templates.
  • [COMMAND_EXECUTION] (HIGH): The workflow relies on executing Python code blocks via the shell (python -c) and running local scripts that are dynamically modified by the agent with potentially tainted data extracted from source documents.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 03:22 AM