ableton-live

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from Ableton Live (track names and MIDI notes) which acts as an external ingestion point for untrusted data.
  • Ingestion points: The skill reads track names via live_get_song_overview in recipes/batch-rename.md and MIDI note data via live_get_notes in recipes/humanize-midi.md.
  • Capability inventory: The skill has the ability to modify projects through tools like live_set_track_props and live_set_notes, and can execute shell commands for configuration.
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between project data and its own system instructions.
  • Sanitization: No validation or filtering is performed on track names or MIDI data before they are processed by the agent's logic.
  • [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to verify the environment and configure the MCP client.
  • Evidence: doctor.md includes an instruction to run node --version to check for prerequisites.
  • Evidence: setup.md provides a command claude mcp add to register the Loophole bridge as an MCP server, incorporating a local authentication token into the command parameters.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 11:52 PM