Modelscope Api
PassAudited by ClawScan on May 6, 2026.
Overview
This looks like a legitimate ModelScope API helper, but users should carefully approve token use, installs, MCP configuration, and any deploy/delete actions.
This skill is reasonable to use for ModelScope discovery and API assistance. Use the least-privileged ModelScope token, prefer setting MODELSCOPE_API_TOKEN instead of pasting tokens into chat, and do not approve deploy/delete/install or MCP configuration changes until you have reviewed the exact command, target, and publisher.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A broad token could allow downloads or write actions in the user's ModelScope account if the user approves those actions.
The skill can use a ModelScope account token. The documentation frames this as session-only and recommends minimal permissions, which is appropriate, but token scope controls what the agent can do.
Token 优先级：1. 环境变量 MODELSCOPE_API_TOKEN ... 2. 直接提供 ... AI 会在本次会话中使用它
Use a read-only or download-scoped token when possible, prefer an environment variable over pasting secrets into chat, and rotate the token if it is exposed.
If approved, the agent may change cloud resources, install components, or delete ModelScope assets depending on the requested action and token permissions.
The skill explicitly includes mutating operations such as deploy, delete, and install, while requiring user confirmation. That is purpose-aligned but still high-impact.
所有变更操作（部署、删除、安装等）需要用户明确确认
Before approving any mutation, verify the exact target, action, token scope, and rollback plan; ask the agent for a dry-run or command preview when possible.
Installing a listed skill or package can run third-party code locally or modify the user's agent environment.
The skill surfaces marketplace installation commands, including remote script execution and package installation. The documentation presents these as user-directed, not automatic.
install_command includes: npx skills add ..., curl -fsSL https://modelscope.cn/skills/install.sh | bash -s -- ..., pip install --upgrade modelscope ...
Install only trusted or verified skills, review source URLs and install commands, and avoid curl-to-bash workflows unless you trust the publisher.
Adding an untrusted MCP server could expose prompts, data, API keys, or tool access through that server's runtime permissions.
The skill retrieves MCP server configurations that can be copied into an agent's MCP client. MCP servers may receive context or expose tools once configured.
server_config | MCP 客户端配置（command + args），直接用于 mcp.json; example command: uvx mcp-server-fetch
Review each MCP server's source_url, verification status, env_schema, and requested credentials before adding it to mcp.json.