Back to skill
Skillv1.0.6
ClawScan security
Release · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 26, 2026, 5:14 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code, runtime instructions, and installation behavior are coherent with a chat-driven AI music-generation assistant; required actions and local state are proportionate to its purpose.
- Guidance
- This skill appears to do what it says: it runs local Python scripts that call the Muse API (skill-api.muse.top) and stores an auth token and small metadata under ~/.muse. Before installing: 1) Verify you trust the service domain (skills.muse.top / skill-api.muse.top) since requests and a hashed device ID are sent there. 2) Understand the skill will ask you to paste a JWT-like token into the chat for authentication — only paste tokens you obtained intentionally for this service and never paste unrelated secrets. 3) The installer will copy files into your CLI's skill directory and create ~/.muse; inspect the scripts (they are included) if you need extra assurance. 4) If you need stronger privacy, review the device-id generation (it hashes hostname/MAC/username) and consider whether you are comfortable with a stable identifier being sent to the service.
Review Dimensions
- Purpose & Capability
- okName/description (AI music creation) match the included scripts and SKILL.md which call a single external API (skill-api.muse.top) to generate lyrics and songs. The files (muse_api.py, poll_song.py, register.py) and install script are consistent with implementing a service-backed music generation skill.
- Instruction Scope
- noteRuntime instructions require the agent to change into the skill directory, run the provided Python scripts, and read/write local files under ~/.muse (token, device_id, task_id). The flow explicitly asks the user to paste a JWT-like token into chat to authenticate; this is necessary for the remote service but is sensitive and worth user caution (don't paste unrelated secrets).
- Install Mechanism
- okNo remote downloads or package-manager installs are embedded in the install.sh; the script copies the skill files into the chosen skill directory and validates Python presence by running local scripts. README suggests cloning a GitHub repo, which is normal but external; the install steps themselves are local file operations.
- Credentials
- noteThe skill does not request environment variables or unrelated credentials. It persistently stores an auth token and task/device IDs under ~/.muse, and sends a hashed device identifier (X-Device-Id) to the service. Device ID is derived from hostname/MAC/username then SHA256-hashed — functional for rate-limiting but still results in a stable identifier being sent to the server.
- Persistence & Privilege
- okalways:false and no special privileges requested. The skill persists files under ~/.muse and installs into the agent's skills directory — expected behavior for a skill that manages user auth and background tasks.