09-track
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script (
build_archive.py) to process and merge analytics data. It also uses thefindcommand to safely locate the script's path within the environment before execution. - [EXTERNAL_DOWNLOADS]: The skill specifies a dependency on the
openpyxlPython library for processing Excel files. This is a standard and well-known package for spreadsheet operations. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including LinkedIn CSV and XLSX exports, creating a potential surface for indirect prompt injection.
- Ingestion points: LinkedIn
AggregateAnalytics(XLSX),Shares.csv,Rich_Media.csv, and local*.mddraft files. - Boundary markers: The skill does not implement specific delimiters for external content, treating it as structured data for archival.
- Capability inventory: The skill writes to a local
archive/directory and executes a local Python script for data processing. - Sanitization: The script uses standard libraries (
csv,json,openpyxl) and regular expressions to parse and clean specific data fields.
Audit Metadata