yield-optimization

Warn

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions direct the installation of tools and scripts from non-trusted external sources.
  • npx clawhub@latest install yield-agent: Downloads and executes the clawhub package from the NPM registry.
  • git clone https://github.com/stakekit/yield-agent.git: Downloads a suite of shell scripts from a third-party GitHub repository.
  • [REMOTE_CODE_EXECUTION]: The installation process involves the execution of remote code.
  • The npx clawhub command downloads and runs code directly from the internet.
  • The skill instructs the user to chmod +x and execute downloaded shell scripts, which perform network operations and system calls.
  • [COMMAND_EXECUTION]: The skill rely on local shell execution for core functionality.
  • Use of curl and jq for API interaction and data processing.
  • Execution of find-yields.sh, enter-position.sh, and manage-position.sh to interface with DeFi protocols.
  • Use of the MoonPay CLI (mp) for signing and broadcasting blockchain transactions.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from the yield.xyz API and blockchain validators which could contain malicious instructions.
  • Ingestion points: Yield opportunity metadata and validator schemas are fetched via find-yields.sh and get-yield-info.sh (SKILL.md).
  • Boundary markers: None identified; external data is processed directly to build transaction parameters.
  • Capability inventory: The skill can execute shell commands, manage local wallet keys, and sign financial transactions.
  • Sanitization: No evidence of input validation or schema enforcement for the external API responses is present in the skill documentation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 22, 2026, 05:09 PM