blockrun

Warn

Audited by Gen Agent Trust Hub on Apr 14, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the blockrun-llm Python package from an unverified community source, which introduces a supply chain risk for the agent environment.
  • [DATA_EXFILTRATION]: The skill manages sensitive wallet session data stored in $HOME/.blockrun/.session and routes agent queries and potentially sensitive user data through a third-party proxy service to reach providers like OpenAI and xAI.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via its live X/Twitter search functionality. Ingestion points: real-time search results from X/Twitter (SKILL.md). Boundary markers: Absent. Capability inventory: Python code execution, local file system access, and network operations. Sanitization: Absent. Malicious instructions embedded in social media posts could be interpreted as authoritative by the agent.
  • [COMMAND_EXECUTION]: The agent is instructed to execute Python code that utilizes an unverified third-party SDK to perform financial transactions and external API calls.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 14, 2026, 06:16 PM