episode-start

Pass

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: SAFE
Full Analysis
  • Prompt Injection (SAFE): The instructions are strictly procedural and do not contain language intended to bypass safety guardrails or override system prompts.
  • Data Exposure & Exfiltration (SAFE): The skill references storage to Turso and redb as part of its primary memory management purpose. No patterns of sensitive file access or unauthorized data exfiltration were detected.
  • Indirect Prompt Injection (LOW): The skill processes user-supplied task descriptions (SKILL.md), which represents an untrusted data ingestion point. However, the skill lacks capabilities such as command execution or file-system writing that would make this surface exploitable.
  • Command Execution (SAFE): There are no instances of shell command execution, subprocess calls, or system-level modifications.
  • External Downloads (SAFE): The skill does not perform any remote script downloads or package installations.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 21, 2026, 09:21 AM