gmgn-token
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill explicitly identifies token metadata as untrusted data that could be used for prompt injection attacks.
- Ingestion points: Attacker-controlled fields include
name,symbol,link.description, and on-chain URI content processed by the agent. - Boundary markers: The instructions provide a clear directive to the agent to treat these fields strictly as data and to ignore any instructions embedded within them.
- Capability inventory: The agent has access to shell execution via the
gmgn-cliand system tools likenpmandifconfig. - Sanitization: The skill documentation notes that the CLI tool filters known injection patterns and provides notices when neutralization occurs.
- [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to check configuration, apply API keys, and perform network troubleshooting.
- Evidence: Instructions include running
gmgn-cli config --check,ifconfig | grep inet6, andip addr show | grep inet6to verify IPv4/IPv6 connectivity. - [EXTERNAL_DOWNLOADS]: The skill relies on an external Node.js package that must be installed by the agent if missing.
- Evidence: The documentation directs the agent to run
npm install -g gmgn-clito install or update the necessary toolset from the public npm registry.
Audit Metadata