rules-creation

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly allows opencode.json to include remote URLs (e.g., "https://raw.githubusercontent.com/..." and "https://example.com/rules.md") and AGENTS.md directs the agent to use its Read tool to load and treat referenced files as mandatory instructions, meaning the agent will fetch and interpret arbitrary public web content that can change its behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). Yes — opencode.json examples show OpenCode will fetch remote instruction URLs at runtime (e.g., https://raw.githubusercontent.com/my-org/public-standards/main/api-design.md) and inject that fetched content into the LLM context as instructions/prompts, creating a runtime dependency that can directly control agent behavior.