prompt-injection-guard

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust security architecture aimed at preventing the execution of malicious instructions derived from untrusted sources (e.g., README files, HTML reports, or scraped content).
  • [SAFE]: It defines a structured 'Provenance x Pattern' decision matrix that ensures high-risk commands are blocked or require explicit user confirmation based on where the instruction originated.
  • [SAFE]: The instructions include patterns for detecting obfuscation techniques such as hidden HTML comments, white-on-white text, and Base64-encoded payloads within processed data.
  • [SAFE]: It lists sensitive system paths (e.g., ~/.ssh/, ~/.aws/) and dangerous shell patterns (e.g., pipe-to-shell) for the purpose of identification and restriction rather than execution.
  • [SAFE]: The skill adheres to security best practices by recommending the use of boundary markers (e.g., INJECTION-RAW blocks) when surfacing suspicious content to the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 05:43 AM
Security Audit — agent-trust-hub — prompt-injection-guard