harness-security-bench

Warn

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses npx to fetch the @metaharness/darwin package from the NPM registry at runtime. While NPM is a well-known service, the package scope is not recognized as a trusted vendor.
  • [REMOTE_CODE_EXECUTION]: The skill executes code downloaded via npx from a remote source. This pattern allows code from an external package to execute in the local environment.
  • [COMMAND_EXECUTION]: The skill constructs and executes shell commands using variable arguments like --population and --cycles. This creates a potential surface for command injection if these arguments are derived from unvalidated user input or configurations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 22, 2026, 06:00 PM
Security Audit — agent-trust-hub — harness-security-bench