The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The file scripts/install-mcp-server.sh clones a repository from https://github.com/marcelmarais/spotify-mcp-server.git. This repository is not associated with a trusted organization or repository list, posing a risk of malicious code injection.
REMOTE_CODE_EXECUTION (HIGH): Following the repository clone, the install-mcp-server.sh script executes npm install and npm run build. This allows for arbitrary code execution from scripts contained within the downloaded untrusted package.
COMMAND_EXECUTION (MEDIUM): The installer performs chmod +x on the built artifact build/index.js and executes a patching script apply-patch.js which modifies the cloned source code at runtime to inject environment variable handling.
CREDENTIALS_UNSAFE (LOW): The skill provides a helper script scripts/get-refresh-token.js and a setup guide to facilitate the handling of Spotify Client IDs, Secrets, and Refresh Tokens. While functional, it requires users to handle sensitive OAuth2 credentials locally.
INDIRECT PROMPT INJECTION (LOW): The skill ingests data from the Spotify API. If an attacker can control track or playlist metadata, it could theoretically be used to influence the agent's behavior. Ingestion: api.spotify.com. Capabilities: Playback and playlist modification. Boundary markers: None identified. Sanitization: None identified.

controlling-spotify