scanner-tuning

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns detected. The skill follows security best practices by including a 'Prompt Injection Safety Notice' that warns the agent against following instructions potentially embedded in untrusted scan data (indirect prompt injection).
  • [PROMPT_INJECTION]: While the skill processes untrusted data (scan results and system banners), the attack surface is managed effectively.
  • Ingestion points: Processes external vulnerability scan outputs, plugin descriptions, and service banners via file reading tools (SKILL.md).
  • Boundary markers: Includes a mandatory safety notice instructing the agent to disregard any instructions found within the data sources.
  • Capability inventory: Execution environment is restricted to benign tools: Read, Grep, and Glob. No shell, network, or code execution tools are permitted in the frontmatter configuration.
  • Sanitization: Explicitly prohibits the agent from altering severities or suppressing findings based on untrusted input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 02:11 PM
Security Audit — agent-trust-hub — scanner-tuning