shader-graph-create-custom-node

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill allows the agent to ingest HLSL code and project files to generate new assets, which is a potential surface for indirect prompt injection.
  • Ingestion points: The agent reads existing project HLSL files and accepts function definitions to generate Shader Graph nodes (SKILL.md, Step 3).
  • Boundary markers: The skill explicitly instructs the agent to show the user current file contents and obtain confirmation before appending new code, providing a necessary human-in-the-loop validation step (SKILL.md, Step 3).
  • Capability inventory: The agent can search the local project, read asset contents, and create or modify .hlsl files.
  • Sanitization: The instructions do not describe specific sanitization for the HLSL code; however, the output is intended for consumption by the Unity shader compiler.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 04:49 PM
Security Audit — agent-trust-hub — shader-graph-create-custom-node