workspace-setup

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly downloads CLAUDE.md/AGENTS.md from a configured Filebrowser server (init_workspace.py, sync_config.py, and filebrowser_client.py) and the skills_manager.py runs npx skills (and references the public skills browsing site https://skills.sh/) to search/install public skill packages, which clearly fetches and ingests untrusted public/user-generated third‑party content that can influence tool execution (e.g., installing/using external skills).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill's runtime scripts (init_workspace.py and sync_config.py) call the configured FileBrowser API (e.g., the configured instance_url such as http://your-server:8080 used with /api/raw{remote_path}) to download CLAUDE.md and AGENTS.md — agent configuration files that the skill fetches at runtime and which can directly control agent prompts/instructions.