chiptune-composer

Warn

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to run local Python scripts such as fix-boundary-gaps.py, merge.py, and extract_handoff.py via shell commands. The script fix-boundary-gaps.py is called with a filename parameter {kebab-case-title}.json which is generated by the agent. If the agent is successfully manipulated by a user's style description, it could generate a malicious filename containing shell metacharacters (e.g., ; rm -rf /), leading to command injection.\n- [REMOTE_CODE_EXECUTION]: The ability to execute arbitrary Python scripts on the host environment represents a significant capability. While the scripts are identified as internal vendor tools (e.g., in music/audio-tracker/tools/), the execution path involves shell interpolation which can be exploited if filename or parameter generation is compromised.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it reads and processes existing song data and metadata to inform its composition.\n
  • Ingestion points: The agent reads music/audio-tracker/songs/index.json and individual song files like boss-battle.json or canon-in-d.json to calibrate its output.\n
  • Boundary markers: The instructions do not specify any delimiters or warnings to ignore potential instructions embedded within these JSON files.\n
  • Capability inventory: The agent has the authority to write new JSON files to the filesystem and execute shell commands via Python.\n
  • Sanitization: No sanitization or validation is performed on the content of the indexed song files or on the input used to generate filenames for the shell commands.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 3, 2026, 07:08 AM