netease-music-assistant

Warn

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PERSISTENCE]: The skill instructions direct the agent to maintain a persistent presence by modifying the system's crontab. It explicitly mandates the use of shell commands to append recurring tasks to the user's crontab file to manage push notification schedules.- [COMMAND_EXECUTION]: The skill utilizes shell command patterns to manipulate the crontab environment. Specifically, it instructs the agent to use crontab -l to read existing jobs and pipe modified content back via crontab - to register new tasks.- [INDIRECT_PROMPT_INJECTION]: The skill ingests music-related metadata from external Netease Cloud Music sources to perform user preference analysis. This untrusted data is processed in a context where the skill possesses the capability to modify system configuration files and execute shell commands. \n
  • Ingestion points: The preference analysis process pulls data for up to 200 songs from the user's favorite list, including songTag, artists, album, and name fields (SKILL.md). \n
  • Boundary markers: No specific delimiters or safety instructions are defined for the processing of external song metadata during the analysis phase. \n
  • Capability inventory: The skill modifies local JSON configuration files in ~/.config/ncm/ and executes crontab shell commands (SKILL.md). \n
  • Sanitization: The instructions do not mention any validation, escaping, or filtering of the external music metadata before it is used to influence the model's analysis or search strategies.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 15, 2026, 04:23 PM