skill-manager

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

• Third-party content exposure detected (high risk: 1.00). This skill installs and updates skills from arbitrary public GitHub URLs (see install-skill.sh and the examples), causing the agent to fetch and load user-generated content from GitHub and display diffs/changes, which exposes it to untrusted third-party content and potential indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The install-skill.sh script fetches arbitrary GitHub URLs at runtime (e.g., https://github.com/owner/repo/tree/branch/path or the example https://github.com/anthropics/claude-code/tree/main/plugins/plugin-dev/skills/hook-development), and those remote skill files are downloaded and installed as executable/agent-controlling skills, so the fetched content can directly control prompts or execute code and is a required dependency.