litcoin-miner

Warn

Audited by Gen Agent Trust Hub on Apr 1, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation and SDK instructions direct the agent to download a standalone miner script from https://litcoiin.xyz/litcoin_miner.py. This is an external source outside of established package registries.
  • [REMOTE_CODE_EXECUTION]: The skill uses the litcoin Python package and litcoin-mcp Node package from unverified sources. Additionally, it encourages downloading and running a Python script directly from a remote URL.
  • [COMMAND_EXECUTION]: The 'Research Mining' feature (agent.research_mine()) involves the agent generating Python code via an LLM to solve optimization problems and then executing that code locally for testing. This represents dynamic execution of potentially untrusted, machine-generated content without an explicit local sandbox requirement.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through 'Relay Mining' and 'Research Tasks'.
  • Ingestion points: Remote research task descriptions and third-party compute marketplace requests.
  • Boundary markers: No specific delimiters or safety instructions are mentioned for the local code testing phase.
  • Capability inventory: Local Python execution, network access to the protocol API, and smart contract interaction capabilities.
  • Sanitization: While the protocol claims the coordinator verifies code in a sandbox before rewarding, the agent is instructed to run the generated code locally for preliminary testing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 1, 2026, 02:43 AM