orchestrate

SUSPICIOUS. The skill’s footprint matches its stated orchestration purpose, but that purpose is inherently high risk: it enables unattended agent-driven code execution and file modification across repos, including third-party CLIs and custom commands. Data flow is mostly local rather than overt exfiltration, so this is not confirmed malware, but it materially increases operational and security risk.

No clear embedded backdoor/stealth exfiltration is visible in this orchestrator script. The primary supply-chain security risk is that it sources agent profile files as shell code ('. "$f"'), allowing arbitrary code execution if an attacker can modify or inject those profile files. Additionally, P_CMD/--cmd are launched to run external agent CLIs, and the script auto-dismisses trust dialogs by sending Enter when configured regexes match. Overall: likely a legitimate orchestration tool with significant attack surface around trusted profile files and command injection via profiles/flags; treat profiles as fully trusted code.