acestep-docs

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation provides instructions to download the project's portable package from the vendor domain files.acemusic.ai. It also references official installers for the uv package manager from the well-known service astral.sh. Additionally, it describes downloading model weights from standard repositories including HuggingFace and ModelScope.
  • [COMMAND_EXECUTION]: The skill provides numerous examples of shell commands and launch scripts (.bat and .sh files) used for environment setup, dependency management, and running the music generation server. These include standard operations for a local development environment such as setting environment variables and executing Python scripts.
  • [INDIRECT_PROMPT_INJECTION]: The system documented by this skill processes user-provided music prompts and lyrics to guide audio generation. This represents an inherent indirect prompt injection surface common to LLM-integrated tools. The documentation suggests the use of structural tags (e.g., <prompt>, <lyrics>) which serve as boundary markers to help distinguish user content from instructions.
  • Ingestion points: User-supplied music descriptions and lyrics provided via API endpoints, Gradio UI fields, and message content (documented in README.md and Openrouter_API.md).
  • Boundary markers: The system uses structural tags like <prompt> and <lyrics> to delimit inputs, as described in Openrouter_API.md.
  • Capability inventory: The skill itself uses Read, Glob, and Grep tools to navigate documentation. The documented system performs ML inference using DiT and LM models.
  • Sanitization: Standard input validation is implied by the API structure, though specific prompt sanitization logic is not detailed in the user-facing documentation.
  • [SAFE]: The update check mechanism described in guides/UPDATE_AND_BACKUP.md is a transparent, git-based system that notifies users of new commits to the project repository. It is a standard practice for open-source local tools and includes user prompts before applying any changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 04:33 PM