music-and-sound
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from video transcripts and timeline metadata to determine music selection and timing.
- Ingestion points: Video timeline windows and transcript text are read from the project context (SKILL.md Sections 1 and 2).
- Boundary markers: The instructions do not define delimiters to separate user-provided data from system instructions.
- Capability inventory: The skill writes to the file system (creating WAV files), executes network requests (curl), and invokes various MCP tools for searching and editing recordings.
- Sanitization: No specific sanitization or filtering of transcript content is mentioned before it is processed for search criteria.
- [COMMAND_EXECUTION]: The skill uses the
curlutility to download audio assets at runtime. - Evidence: Instructions direct the agent to fetch signed URLs provided by the Epidemic Sound MCP server into local directories.
- Context: This behavior is limited to the primary purpose of retrieving audio files and does not appear to facilitate arbitrary command execution or exfiltration.
Audit Metadata