experiment-workspace

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local statistical analysis and data collection scripts using python and npx tsx. These operations are well-documented and are core to the skill's purpose of processing experimental data.
  • [EXTERNAL_DOWNLOADS]: Instructions include installing standard, well-known numerical libraries (numpy, scipy) via the system package manager (pip). These are trusted, industry-standard packages.
  • [DATA_EXFILTRATION]: The skill interacts with an external database API to store and retrieve experiment records. Security is maintained by instructing users to manage credentials (like API tokens and database URLs) through environment variables rather than hardcoding them within the skill.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests experimental metric data from a database. This data is handled as structured JSON and used strictly for statistical calculations, presenting a minimal surface for instruction-based attacks compared to skills that process unstructured natural language from untrusted sources.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 06:45 AM
Security Audit — agent-trust-hub — experiment-workspace