harness-learn

Warn

Audited by Snyk on Jul 6, 2026

Risk Level: MEDIUM
Full Analysis

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (high risk: 1.00). The skill requires cloning and invoking the external metaharness repository at https://github.com/ruvnet/metaharness.git (git clone shown), and that upstream metaharness code (metaharness learn) executes runtime code and evolves policy prompts, so the fetched repo directly controls prompts/executes code.

Issues (1)

W012
MEDIUM

Unverifiable external dependency detected (runtime URL that controls agent).

Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 05:39 PM
Issues
1
Security Audit — snyk — harness-learn