Skills
skills/starchild-ai-agent/official-skills/mantle/Gen Agent Trust Hub

mantle

Fail

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill installs the mantle-cli utility directly from the official Mantle GitHub organization (github:mantle-xyz/mantle-agent-scaffold).
  • [COMMAND_EXECUTION]: Extensively uses shell commands via npx and mantle-cli to perform blockchain queries, build transactions, and interact with the Mantle registry.
  • [REMOTE_CODE_EXECUTION]: Installation of packages from remote GitHub repositories via npm constitutes a remote code execution vector, though it targets the official ecosystem repository.
  • [DATA_EXFILTRATION]: The mantle-data-indexer skill allows the agent to send GraphQL and SQL queries to user-provided external endpoints, which is a functional requirement for indexing but presents a potential route for data transmission.
  • [PROMPT_INJECTION]: The skill processes external data from blockchain indexers and transaction simulation backends, representing an indirect prompt injection surface. The skill mitigates this with strict output templates and boundary markers for unverified data.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Apr 10, 2026, 12:48 PM