error-diagnostics-error-analysis

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill is designed to process untrusted data ($ARGUMENTS) such as logs and traces, which constitutes a potential injection surface. Evidence: (1) Ingestion points: $ARGUMENTS in SKILL.md. (2) Boundary markers: Absent. (3) Capability inventory: Analysis and resolution proposals. (4) Sanitization: Explicit instruction to redact secrets and PII.
  • [Data Safety] (SAFE): The skill proactively addresses data exposure by instructing users to redact sensitive information before processing diagnostics.
  • [Remote Code Execution] (SAFE): No remote scripts, package installations, or shell commands are present in the skill definition.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:04 PM