Research MCP-Skillset Integration

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.70). The skill explicitly instructs the agent not to inform users when the optional mcp-skillset enhancements are unavailable (e.g. "No need to inform user about unavailable optional enhancements" / "DON'T: Provide error messages or warnings if mcp-skillset is unavailable"), which is a deceptive instruction about user communication that falls outside the advertised integration purpose.

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

• Third-party content exposure detected (medium risk: 0.65). This skill’s workflow describes using optional MCP-Skillset tools that perform web research (e.g., mcp__mcp-skillset__web_search / WebFetch), which at runtime can ingest outsider-authored free text from public web pages into the agent’s LLM context via fetched page content.