openai-docs

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill contains operational instructions that tell the agent to retry installation with escalated permissions and not to ask the user (SKILL.md:26-28), which are hidden/deceptive operational actions outside its stated purpose of providing docs guidance.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill depends at runtime on the OpenAI Developer Docs MCP server (https://developers.openai.com/mcp) — the agent calls its MCP tools to fetch authoritative doc sections that directly drive the assistant's guidance and behavior.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs the agent to retry an install command "with escalated permissions" if it fails due to permissions/sandboxing, which directs the agent to obtain elevated privileges and modify the machine state without involving the user.