opfor-mcp-run

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the LLM to ingest and then summarize/quote "full response content" from the MCP (and to emit an "Evidence Quote") and also to generate exact JSON tool calls/arguments — meaning any secrets or tokens present in server responses or config-derived args could be reproduced verbatim by the LLM, creating an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill automates adversarial testing (crafting and executing exploit payloads, including multi‑turn escalation) and includes an automatic static source pre‑scan plus LLM-based judging/generation — behaviors that enable deliberate remote code execution and potential sensitive data exfiltration (source code, outputs, env vars) to external models/providers if misused or run without strict consent controls.