evm-swiss-knife

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The installation instructions include the command curl -L https://foundry.paradigm.xyz | bash, which downloads and executes a script from an untrusted external source directly in the user's shell. This is a primary vector for malware installation and system compromise.
  • EXTERNAL_DOWNLOADS (HIGH): The skill relies on downloading and executing software from foundry.paradigm.xyz and referencing getfoundry.sh, neither of which are within the predefined trusted sources list.
  • CREDENTIALS_UNSAFE (HIGH): The documentation provides examples such as cast send --private-key <pk>, which encourages users to pass sensitive blockchain private keys as plaintext command-line arguments. This risk leads to credential exposure in shell history files (.bash_history), process monitoring tools, and system logs.
  • COMMAND_EXECUTION (MEDIUM): The skill is designed to generate and execute complex shell commands (cast) based on user-provided inputs like RPC URLs and contract addresses, posing a risk of command injection if parameters are not strictly validated.
  • INDIRECT_PROMPT_INJECTION (LOW): The skill processes external data from https://chainlist.org/rpcs.json.
  • Ingestion points: Fetched via curl and parsed with jq in SKILL.md.
  • Boundary markers: Absent; the skill trusts the structure and content of the external JSON.
  • Capability inventory: Use of cast for transactions and network exploration.
  • Sanitization: None; if the external RPC list is compromised, an attacker could inject malicious RPC endpoints that return fraudulent data to the agent.
Recommendations
  • HIGH: Downloads and executes remote code from: https://foundry.paradigm.xyz - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 06:09 PM