agent-wallet
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructs the agent to handle sensitive passwords and API credentials insecurely. It documents the
--save-runtime-secretsflag, which saves the master password in unencrypted plaintext to~/.agent-wallet/runtime_secrets.json. It also suggests asking users for passwords or Privy API secrets in conversation and passing them as command-line arguments, exposing them to system-level process monitoring. - [DATA_EXFILTRATION]: The skill targets sensitive directories and files used for wallet storage, specifically the
~/.agent-walletdirectory and thewallets_config.jsonandruntime_secrets.jsonfiles. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied JSON payloads for transaction and message signing, creating an injection surface. Ingestion points: The
signcommands inSKILL.mdinterpolate user-provided strings directly into CLI commands. Boundary markers: There are no markers or safety instructions to prevent the agent from interpreting commands embedded within these payloads. Capability inventory: The CLI can perform signing operations for EVM and TRON networks. Sanitization: The skill provides guidance on shell-safe quoting but lacks validation or sanitization of the signing payload content itself. - [COMMAND_EXECUTION]: The skill dynamically constructs and executes shell commands using the
agent-walletCLI with parameters and payloads derived from user input. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
@bankofai/agent-walletpackage from the NPM registry.
Audit Metadata