data-importer
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from CSV, Excel, and Shopify exports, which provides a surface for indirect prompt injection where malicious instructions embedded in product names or attributes could influence agent behavior.
- Ingestion points: External files (CSV, .xlsx, Shopify exports) referenced in SKILL.md and shopify-format.md.
- Boundary markers: None explicitly defined to separate data from instructions.
- Capability inventory: File system reading, shell command execution (tr, sed, column), and Python script execution.
- Sanitization: Basic cleaning for slug generation is mentioned but not specifically for preventing injection attempts.
- [COMMAND_EXECUTION]: The skill provides shell command snippets in SKILL.md and references/csv-patterns.md that use variable interpolation ($FILE) without explicit escaping instructions. If the agent executes these commands by directly substituting a filename containing shell metacharacters, it could lead to arbitrary command execution.
- [DYNAMIC_EXECUTION]: The skill suggests using python3 -c to execute a Pandas-based script for Excel to CSV conversion. This runtime generation and execution of code based on external file paths introduces a risk if the input path is not properly validated or escaped before execution.
Audit Metadata