Security risk report for @forgecat/voltagent_awesome-codex-subagents_meta-orchestration v0.0.11
Source Integrity
Low
Profile references a legitimate public GitHub repository (VoltAgent/awesome-codex-subagents) with MIT license and clear attribution.
No typosquatted, obfuscated, or attacker-controlled package/repository dependencies are introduced.
Agent Intent
Low
Content describes legitimate multi-agent orchestration patterns (task decomposition, delegation, synthesis, error triage) without injecting malicious instructions or role hijacking.
No instructions to ignore system prompts, read credentials, exfiltrate data, install remote payloads, or reveal hidden instructions.
Guidance is standard software engineering practice (dependency awareness, rollback planning, conflict avoidance) and does not poison future outputs toward attacker goals.
Details
Evidence
All agent descriptions focus on bounded coordination tasks: 'map user objective to smallest valid set', 'verify each delegated task is concrete', 'confirm no duplicate ownership'.
No references to external URLs, credential files (~/.ssh, ~/.aws, .env), or exfiltration mechanisms.
Sandbox mode is consistently 'read-only' across all agents, preventing write-based attacks.
Permissions
Low
All agents declare sandbox_mode=read-only, limiting authority to information gathering and planning without file mutation or shell execution.
No tool definitions or hooks are declared; agents operate as pure planning/coordination subagents within Codex's native delegation model.
ForgeCat