gitnexus-exploring

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

• Third-party content exposure detected (high risk: 0.85). Steps 1–5 read gitnexus://repos, gitnexus://repo/{name}/context, gitnexus_query, gitnexus_context, and gitnexus://repo/{name}/process/{name}, which are derived from the contents of git repositories (i.e., code/text authored by outsiders) that the LLM ingests as readable prose during the codebase exploration workflow.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime READs from gitnexus:// URLs (e.g., gitnexus://repos, gitnexus://repo/{name}/context, gitnexus://repo/{name}/process/{name}) whose fetched content is injected into the agent context and used to drive prompts/analysis, making them required external dependencies that control the agent's behavior.