ltv-predictor

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE] (SAFE): No malicious behavior detected. The skill is designed for ecommerce and retail analytics.
  • [Data Processing] (SAFE): The tool processes local CSV data files using established libraries like pandas and scikit-learn. It performs feature engineering and regression modeling as documented without suspicious data access or exfiltration.
  • [Dependencies] (SAFE): All identified dependencies (pandas, scikit-learn, etc.) are standard, verifiable packages from the PyPI registry. The inclusion of Flask is consistent with the documented ability to start a local API server for predictions.
  • [Indirect Prompt Injection] (SAFE): Although the skill ingests untrusted CSV data which constitutes a theoretical attack surface, its primary purpose is numerical regression rather than feeding text back into an LLM prompt. No exploitable interpolation patterns were found in the provided examples.
  • [Capabilities] (SAFE): The permissions requested in SKILL.md (Bash, Write, Read) are appropriate for its core functions, such as installing packages, executing training scripts, and generating report files (HTML, Excel, Markdown).
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:12 PM