output-sanitizer

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious patterns or behaviors were detected; the skill acts as a defensive security utility.
  • [NO_CODE]: The skill contains only instructional markdown and lacks executable code or dependencies.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted agent output.
  • Ingestion points: Agent responses processed by the module (SKILL.md).
  • Boundary markers: No explicit data delimiters are defined in the instructions.
  • Capability inventory: Permission 'fileRead' is defined in metadata but is not utilized in the skill's logic.
  • Sanitization: Implements regex-based redaction for various secret types and PII.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 04:15 PM