optimize-agent-prompt

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

• Potentially malicious external URL detected (high risk: 0.90). The skill states the Agent Context Document is "injected automatically via the MCP server" and provides an MCP URL template (https://api.sanity.io/vX/agent-context/PROJECT_ID/DATASET/SLUG), so content fetched from that URL at runtime directly supplies the agent's instructions/prompts.