trade-prediction-markets

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: Potential surface for indirect prompt injection via untrusted market data. The skill describes a workflow where data from Polymarket events is ingested and used as context for generating trading code.
  • Ingestion points: The mcp__workbench__get_all_prediction_events and mcp__workbench__get_prediction_market_data tools retrieve event names and metadata from external sources (SKILL.md).
  • Boundary markers: None mentioned. The workflow does not describe the use of delimiters or instructions to ignore embedded commands in the ingested market data.
  • Capability inventory: The mcp__workbench__create_prediction_market_strategy tool generates executable Python code, which could be influenced by malicious instructions in market metadata.
  • Sanitization: No evidence of validation or sanitization for the strings retrieved from the prediction market APIs.
  • [NO_CODE]: The skill package consists exclusively of the SKILL.md markdown file. It contains no source code, binaries, or executable scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 07:00 AM