The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by accepting and processing user-provided arguments.
Ingestion points: The $ARGUMENTS variable in SKILL.md is used to ingest an optional album name from the user.
Boundary markers: There are no explicit delimiters (e.g., XML tags or triple quotes) or specific instructions to the model to ignore embedded commands within the input.
Capability inventory: The skill is authorized to use Read, Glob, and Grep tools for file system access, as well as the specialized bitwize-music-mcp toolset for music project management.
Sanitization: The skill performs a 'fuzzy match' on the user input but does not implement validation or escaping to prevent the input from influencing the agent's logic flow beyond identifying an album.

next-step