The Agent Skills Directory

Remote Code Execution (HIGH): The skill suggests running curl -LsSf https://astral.sh/uv/install.sh | sh in scripts/kernel.py to install the uv tool. This piped shell execution pattern is a high-risk security vector.\n- External Downloads (MEDIUM): The install_packages method in scripts/kernel.py allows the installation of any Python package via uv pip install, which could be exploited to load malicious code.\n- Command Execution (MEDIUM): The skill executes arbitrary Python code in a persistent Jupyter kernel as part of its core analysis functionality.\n- Prompt Injection (LOW): The skill is vulnerable to indirect prompt injection. 1. Ingestion points: User business questions and database results (SKILL.md). 2. Boundary markers: No delimiters isolate external data from instructions. 3. Capability inventory: Python execution, SQL execution, and package installation. 4. Sanitization: No validation or sanitization is performed on external inputs.

analyzing-data