factor-research

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests instrument factor values and forward returns from external CSV files via the factor_csv and return_csv parameters. This creates a potential surface for indirect prompt injection if the files were to contain malicious instructions instead of expected numeric data.
  • Ingestion points: Path parameters for factor and return data in the factor_analysis tool and various quantlib module calls.
  • Boundary markers: The documentation does not specify the use of delimiters or specific safety instructions to isolate ingested data from the agent's prompt context.
  • Capability inventory: The skill is capable of writing research artifacts to a user-defined output_dir and performing complex statistical operations through the quantlib_call tool.
  • Sanitization: While the skill assumes valid aligned CSV structures (date index and instrument columns), there is no mention of explicit content sanitization for the data contained within the files.
  • [DYNAMIC_EXECUTION]: The framework utilizes a quantlib_call mechanism to dynamically load and execute specific mathematical modules for factor combination and neutralization.
  • Evidence: Files references/combination.md and references/neutralization.md describe using a JSON-based envelope to call internal functions such as combine_factors and neutralize_factor.
  • This represents a controlled and limited use of dynamic execution, as calls are restricted to pre-defined, platform-specific research modules rather than arbitrary user-supplied code.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 12:04 PM