dev-guides-navigator

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md “Core Workflow” guide-search (modes 1 & 2), the runtime fetches and then applies outsider-provided text from the published catalogs and guide bodies (e.g., llms.txt, topic index.md, and specific raw GitHub .md) into the main conversation context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches live indexes and raw markdown at runtime from https://camoa.github.io/dev-guides/ (e.g. llms.txt, agentic-recipes.txt, process-recipes.txt) and corresponding https://raw.githubusercontent.com/camoa/dev-guides/... raw files, and those fetched bodies are applied by the agent to drive routing and to apply recipe/guide behavior, so remote content directly influences the agent's instructions and behavior.