simota-agent-skills

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow is installing/invoking a third-party package (simota/agent-skills) from a public registry (GitHub via npx), and at runtime the agent will load that package’s code/content which may include outsider-authored free text that the LLM ingests (e.g., bundled prompts/templates or skill descriptions).

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

• Potentially malicious external URL detected (high risk: 0.90). The install command ("npx skills add simota/agent-skills") fetches and installs content from the referenced GitHub repo (https://github.com/simota/agent-skills), which at runtime would supply skill definitions/prompts that directly control the agent's behavior, so this is a runtime remote dependency that can change agent prompts.