askprisma
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted business data from local files and SQL databases, creating a surface for indirect prompt injection. \n
- Ingestion points: SKILL.md defines phases to scan the working directory for CSV, Excel, Parquet, and JSON files, and to connect to SQL databases. \n
- Boundary markers: The skill uses a structured workflow (Discovery -> Planning -> Iterative Execution) and requires profiling data for raw facts before synthesis. \n
- Capability inventory: The agent can execute Python scripts via Bash, install packages via pip, and write report artifacts to the filesystem. \n
- Sanitization: scripts/generate_report.py implements HTML sanitization for content included in generated PDF reports. \n- [DYNAMIC_EXECUTION]: The skill dynamically generates Python code to perform data analysis tasks and visualization, which is subsequently executed via the shell. \n- [COMMAND_EXECUTION]: The skill uses shell commands for directory creation, execution of analysis scripts, and installing missing Python dependencies. \n- [EXTERNAL_DOWNLOADS]: The skill enables the installation of missing Python libraries from standard package registries via pip install.
Audit Metadata