demoscene-coding

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWNO_CODE
Full Analysis
  • Indirect Prompt Injection (LOW): The skill grounds its behavior in external files, creating an ingestion surface for untrusted data. * Ingestion points: references/patterns.md, references/sharp_edges.md, and references/validations.md. * Boundary markers: Absent; the prompt does not define delimiters to isolate reference content. * Capability inventory: None; no executable scripts or system-level tools are defined. * Sanitization: Absent; there is no logic to validate or escape the content of the referenced files.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 03:15 AM