concepts

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow only fetches its own bundled reference markdown files from the specified GitHub raw URLs (outsider-authored web content), which are then ingested as readable text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to fetch remote markdown at runtime (e.g. https://raw.githubusercontent.com/therealharpaljadeja/monskills/main/concepts/references/async-execution.md and other raw.githubusercontent.com references), which would be injected into the agent's context and thus can directly control prompts/instructions.