shader-fills

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the command pnpm exec tsx apply-shader.ts to execute a local script for modifying HTML files. This operation is confined to the local project environment and uses standard developer tools.\n- [PROMPT_INJECTION]: The skill processes untrusted data from existing HTML files and user-defined descriptions to generate shader code, creating a surface for indirect prompt injection.\n
  • Ingestion points: The skill reads HTML source via read-source-file and accepts natural language descriptions for custom shaders.\n
  • Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the read HTML content.\n
  • Capability inventory: The agent has capabilities to read/write local files (read-source-file, apply-source-edit) and execute local TypeScript files.\n
  • Sanitization: The skill implements validateShaderDef() to ensure that generated output does not contain sequences like </script> or */ that could break the HTML structure or lead to script injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 04:51 AM
Security Audit — agent-trust-hub — shader-fills