metal-shader-expert
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEOBFUSCATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The file
reference.mduses HTML entities (<,>) to represent comparison operators and template brackets in Metal Shading Language (MSL) code samples. - Evidence: Code snippets such as
texture2d<float>and logic likei < debug_countuse escaped characters. While likely a result of markdown rendering, these encodings can be used to obfuscate intent. The decoded content is legitimate MSL syntax. - [INDIRECT_PROMPT_INJECTION]: The skill is configured to use web-research tools (
WebFetch,Firecrawl) to ingest external papers and documentation, which could contain malicious instructions. - Ingestion points: External content fetched via
WebFetchandmcp__firecrawl__firecrawl_search(SKILL.md). - Boundary markers: Absent; no instructions provided to the agent to delimit or ignore instructions in fetched data.
- Capability inventory: The skill can modify the file system (
Write,Edit) and execute specialized shell commands for Metal compilation (xcrun,metal,metallib). - Sanitization: Absent.
Audit Metadata