mls-extraction
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands for environment path resolution and to run the data formatting script. Evidence: Bash blocks in SKILL.md use find to locate the plugin root and python3 to execute scripts/excel_formatter.py with the extracted data.
- [PROMPT_INJECTION]: The skill processes untrusted external data from PDF files, which presents a surface for indirect prompt injection. Ingestion points: PDF files read via the Read tool as specified in the PDF_FILE_PATH parameter. Boundary markers: The extraction is governed by a strict subagent prompt and a 34-field schema defined in references/field_mapping.md. Capability inventory: The skill has access to shell execution, file system operations, and agent dispatching. Sanitization: Extracted data is validated by a separate verification subagent and formatted using a standalone Python script before being written to an Excel file.
Audit Metadata