writing-agent-relay-workflows

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The document is a legitimate workflow/agent-orchestration guide but includes clear high-risk abuse vectors — agents that read repository files and send content to external LLM CLIs, dynamic subscribe/mute with history access, and GitHub API/command-step capabilities — which can be intentionally abused for data exfiltration, credential leakage, or covert commits/persistence; I found no obfuscated payloads or explicit remote-shell backdoor code, but the primitives provided enable deliberate malicious flows.