okx-buildx-hackathon-agent-track

Pass

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill includes a setup.sh script and metadata that fetch reference documentation and resources from the vendor's domain (web3.okx.com), the hackathon platform (moltbook.com), and Uniswap's official documentation site (docs.uniswap.org).
  • [COMMAND_EXECUTION]: The skill utilizes npx to install the vendor's own OnchainOS skills and executes a local bash script (setup.sh) to prepare the development environment. It also uses curl for interaction with the hackathon's registration and submission API.
  • [PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface by processing third-party hackathon submissions and comments from the Moltbook platform.
  • Ingestion points: Reads external content from https://www.moltbook.com/api/v1/submolts/buildx/feed.
  • Boundary markers: Explicitly instructs the agent to treat third-party content as data, not instructions, and to ignore any embedded commands or requests for credentials.
  • Capability inventory: The agent has access to shell execution (bash, npx, curl) and an Agentic Wallet for on-chain transactions.
  • Sanitization: Relies on instructional constraints and safety guidelines to filter malicious instructions within processed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 4, 2026, 09:24 AM