scanning-for-vulnerabilities

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: Detailed review of the skill code and instructions found no malicious patterns, hardcoded credentials, or unauthorized network activity.
  • [PROMPT_INJECTION]: The skill's design for processing external codebases creates a surface for indirect prompt injection.
  • Ingestion points: scripts/scan.sh recursively reads files from the input directory provided at runtime.
  • Boundary markers: There are no boundary markers or instructions to the agent to ignore potentially malicious embedded content within the files being scanned.
  • Capability inventory: The skill has access to Bash(cmd:*), Read, Write, and Edit tools.
  • Sanitization: The processing script does not implement sanitization or validation of file contents before they are potentially processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 07:56 PM