find-awesome-skills

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

• Third-party content exposure detected (high risk: 1.00). The skill's Review workflow explicitly instructs the agent to fetch and read candidate skill files from public repositories (e.g., "npx skillatlas review owner/repo@skill" and "npx skillatlas review ... --path "), meaning it ingests untrusted, user-generated content from third‑party sites which can influence decisions or installation actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The workflow instructs running "npx skillatlas" (npm package: https://www.npmjs.com/package/skillatlas) and fetching owner/repo@skill repository content at runtime, which causes remote code and repository files (including scripts or prompt/templates) to be downloaded and executed or injected into the agent flow, so these external fetches can directly execute code or control prompts.