{
  "handle": "AgentSandbox",
  "mechanism": "Execution isolation for AI {{agent}}s treating them as untrusted insiders. The {{agent}} runs in a containerized environment (gVisor, Firecracker microVM) with: resource quotas (CPU, memory, time), network egress allowlists, filesystem restrictions, and comprehensive logging. Code generated by the {{agent}} is executed in this {{sandbox}}, preventing impact on production systems even if the agent is compromised or generates malicious code. It wraps the execution {{context}} with {{input_guard}} to sanitize ingress data and {{output_guard}} to prevent exfiltration, effectively placing the {{agent}} in a digital quarantine.",
  "gloss": "Isolate agent execution with resource limits and egress controls",
  "failure_modes": [
    "{{sandbox}} Escape: Exploiting container vulnerabilities.",
    "Quota Starvation: Legitimate tasks failing due to overly restrictive limits.",
    "Log Blindness: Missing critical events in {{audit}} logs."
  ],
  "invariants": [
    "Network egress is whitelist-only.",
    "The sandboxed process cannot reach the host filesystem.",
    "Declares resource quotas for CPU, memory and time.",
    "Every action inside the {{sandbox}} is logged."
  ],
  "preconditions": [
    "Resource limits defined",
    "Untrusted code/agent"
  ],
  "postconditions": [
    "Execution completed or terminated by supervisor"
  ],
  "_meta": {
    "tier": 1,
    "ring": 0,
    "related": [
      "Solution",
      "Task"
    ],
    "supersedes": [
      "sema:AgentSandbox#mh:SHA-256:ce7e2d28fa6aefffa0164d17cebc0266c4f7e422cdeb0a7043e078d5469c4be9",
      "sema:AgentSandbox#mh:SHA-256:06f29db915e1bbe17ebf20bab2964279a58c949bc61d386ab20de84749331fe6",
      "sema:AgentSandbox#mh:SHA-256:c092338ebd9e7025254ae45974d7a9854b21b12e5d2943f58f4c0e920f5e1fde",
      "sema:AgentSandbox#mh:SHA-256:d1da527c28217842fbdb7333e85fb95d70c54672ed8a2f37325cde3c5d32d5a5",
      "sema:AgentSandbox#mh:SHA-256:41df2a0465da388dd7ecc55842ea08850e6d9e069c47c6d93e80ba13a0445084"
    ],
    "path": [
      "Society",
      "Protocols"
    ]
  },
  "sema_id": "sema:AgentSandbox#mh:SHA-256:1594a10a882956e4b8501764091e1d5025ba0b405cce34b7ef931bb17e5c634a",
  "sema_ref": "AgentSandbox#1594",
  "sema_stub": "1594",
  "signature": [
    "Sandbox(Agent)"
  ],
  "dependencies": {
    "composes_with": {
      "input_guard": "sema:InputGuard#mh:SHA-256:0ba715e91627fa1f6759a739da0d4159cac6404c8b45b6e97910d98ef0e8a079",
      "output_guard": "sema:OutputGuard#mh:SHA-256:3618ee4d6aa243c10c16af73e4a385a6fa0c60ee193edb0a67d23e2c8d57f93e"
    },
    "references": {
      "agent": "sema:Agent#mh:SHA-256:e127459069739fffc96a71d54ef6e8c97ca53471d5349c16aa72a092a33756cd",
      "audit": "sema:Audit#mh:SHA-256:be970add829f4f162533569b04c9ffd12ea4d070d53405a4993f0785d4074f77",
      "context": "sema:Context#mh:SHA-256:c1b72256dc4732033ee42cc07ff5d9fb60a7f1a3e5634449c343692a33340756",
      "sandbox": "sema:Sandbox#mh:SHA-256:7106e315b4bf7c5bd418fbab5c18e54be940367f77b8bde4756d1cb95beaa16b"
    }
  },
  "sema_layer": "Society",
  "sema_category": "Protocols",
  "used_by": {
    "references": {
      "EmpathySim": "sema:EmpathySim#mh:SHA-256:33c4c1083c17544686139953481e1467b780ec974093cbe6831cc710a5621d0a",
      "MentalSim": "sema:MentalSim#mh:SHA-256:c28c84db7597a2e8cef38b73dc9780b6f8fe28c9d45700c4431b3a7a2ac3b8fc"
    }
  }
}