project-skill-installer

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly searches public/community registries (via npx skills find targeting the skills.sh community registry) and reads SKILL.md files from discovered skill packages as part of its discovery and verification workflow, so untrusted third-party skill descriptions/markdown could be ingested and influence recommendations or install actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes npx commands at runtime (e.g., "npx skills find/add ..." and "npx @tiktok-fe/skills find/add ..."), which fetch packages from the npm registry (e.g., https://registry.npmjs.org/) and will download/execute remote code and can supply SKILL.md content that controls agent behavior—these are required for discovery/installation.