project-references

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

• Third-party content exposure detected (high risk: 0.80). The required workflow can clone/pull from GitHub repos listed in ~/claude-shared/projekte.txt or discovered via gh repo list, after which the LLM reads outsider-authored repository files via cat/grep/find (free text from those repos enters the LLM context), creating an indirect prompt-injection path.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runs git/gh at runtime to clone or pull external repositories (e.g., "owner/repo", "owner/other-service", "owner/your-repo") and reads their files to drive its responses, so fetched remote repository content can directly control agent prompts.