python-workspace

Pass

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute Windows-based Python scripts and the MetaTrader 5 compiler (metaeditor64.exe) through the Wine compatibility layer. Commands are primarily used for market data export and CLI-based indicator compilation within the CrossOver bottle environment.
  • [DYNAMIC_EXECUTION]: The skill facilitates the compilation of MQL5 source code into executable .ex5 files using the official MetaQuotes compiler. It also executes local Python scripts to generate configurations and perform data validation against reference exports.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it reads and processes external MQL5 source code and CSV market data while possessing execution capabilities.
  • Ingestion points: Reads .mq5 source files for translation and .csv files for validation (e.g., in validate_indicator.py).
  • Boundary markers: The instructions do not specify any delimiters or warnings to prevent the agent from following instructions potentially embedded within the data or source code being analyzed.
  • Capability inventory: Full access to Bash and Write tools allows for shell execution and filesystem modification.
  • Sanitization: No explicit sanitization or filtering of external input content is mentioned.
  • [PERSISTENCE]: The skill includes instructions for 'Self-Evolving' behavior, where the agent is prompted to modify the SKILL.md file directly to fix errors, update parameters, or document workarounds. This allows the agent to persistently alter its own instructions and behavior based on output from its execution environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 18, 2026, 02:08 PM