backtest-diagnose

Warn

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill guides the agent to use the bash tool to resolve system errors and perform code validation. Specifically, it suggests running pip install for missing dependencies and executing Python one-liners for syntax checking.
  • [EXTERNAL_DOWNLOADS]: The instruction to use bash("pip install xxx") to fix ImportError allows the agent to download and install arbitrary packages from external registries, which could lead to the installation of malicious code if the package name is manipulated.
  • [INDIRECT_PROMPT_INJECTION]: The diagnostic workflow involves reading multiple external files that could be controlled by an attacker or contain untrusted content. This creates a surface where embedded instructions in these files could subvert the agent's logic.
  • Ingestion points: artifacts/metrics.csv, artifacts/equity.csv, artifacts/trades.csv, code/signal_engine.py, config.json, and run_card.json are all read into the agent's context.
  • Boundary markers: The instructions do not define delimiters or specific markers to separate these file contents from the agent's internal logic.
  • Capability inventory: The agent has access to bash for command execution, edit_file for source code modification, and refresh_strategy_evidence for state updates.
  • Sanitization: No evidence of input validation or sanitization is present before the data is used to decide on code edits or shell commands.
  • [DYNAMIC_EXECUTION]: The skill uses a shell command to execute Python code at runtime (python -c "import ast; ...") to validate the Abstract Syntax Tree (AST) of the signal_engine.py file after it has been edited.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 6, 2026, 02:20 PM