scorestarling-notation

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including PDFs, images, and various music file formats (MIDI, MusicXML, etc.), to create and edit musical scores. This represents an indirect prompt injection surface where malicious instructions could be embedded within the music data to influence the agent's behavior.
  • Ingestion points: User-provided files (PDF, scans, MusicXML, MIDI, ABC, Guitar Pro) and music descriptions provided in the chat.
  • Boundary markers: The "Shared MCP essentials" section provides general guidance on using delimiters for external content, though specific implementation details for the music notation parsing are not provided.
  • Capability inventory: The skill utilizes tools such as create_score, revise_score, edit_score, and unlock_score which interact with a remote MCP server to perform music-related tasks.
  • Sanitization: No explicit sanitization or validation of the semantic content of the music files is mentioned beyond standard tool processing and agent-led manual review steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 03:58 AM