skills/afelipeg/anthropic-skills-for-enterprise-marketing-os/audience-segmentation-brief/Gen Agent Trust Hub
audience-segmentation-brief
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: Detailed analysis shows no security issues. The skill performs intended data processing tasks locally within the provided scripts.
- [COMMAND_EXECUTION]: The skill uses Python scripts (
segmentation_engine.py,segment_overlap_checker.py, andaddressability_checker.py) for data analysis. These scripts are restricted to mathematical and logical operations on input data and do not execute arbitrary shell commands or access sensitive system resources. - [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection via the processing of untrusted customer data. Mandatory evidence chain: 1. Ingestion points:
segmentation_engine.pyandaddressability_checker.pyingest JSON formatted transaction and contact data. 2. Boundary markers: The output templates inSKILL.mdinterpolate data into markdown without specific delimiters. 3. Capability inventory: Scripts are restricted to local processing and do not have network or subprocess capabilities. 4. Sanitization: Input data is normalized for matching (e.g., lowercasing) but is not escaped for the final display. The risk is negligible due to the lack of exploitable tools.
Audit Metadata