fix-sphinx

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md:9-10 and 55-57 describe reading a user-provided GitHub issue URL and analyzing its issue details; that issue body is outsider-authored free text fetched at runtime and then used in the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly looks for a GitHub issue URL pattern and instructs the agent to "Read the issue" (https://github.com/Azure/azure-sdk-for-python/issues/...) at runtime, meaning it would fetch remote issue content that directly guides the agent's actions.