Loading ToolBench
Production-grade enterprise multi-tenant agent sandbox security guard. Enforces strict memory boundary separation and prevents cross-tenant credential leakage.
Single tool 'enforce_tenant_isolation' has critical definition gaps. Tool name lacks action verb (should be 'verify_tenant_isolation' or 'check_tenant_isolation'). Description is present (89 chars) but generic and lacks WHEN/WHY context. Input schema exists but is severely incomplete: execution_payload.cross_tenant_token has type 'null' (invalid JSON Schema), properties lack required array, and no output schema is documented. Parameters lack actionable constraints. No error handling guidance. Tool appears to be a security control but definition does not convey this clearly to LLMs.
Enforces cryptographic multi-tenant sandbox boundaries for enterprise agents.
Tool name lacks action verb. 'enforce_tenant_isolation' is a noun phrase; should start with verb like 'verify_', 'check_', or 'validate_' to signal intent to LLM.
Input schema has invalid type 'null' for cross_tenant_token field. JSON Schema does not support null as a type value; should be {"type": ["string", "null"]} or omitted if unused.
No output schema documented. LLM cannot predict response structure (sandbox_status, tenant_scope_token, etc.). Must document all response fields with types and descriptions.
Description lacks WHEN/WHY context. Does not explain when to call this tool vs alternatives, what prerequisites exist, or what the return values mean. Current text: 'Enforces cryptographic multi-tenant sandbox boundaries for enterprise agents.', too vague.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
No error handling guidance. Tool can return 'QUARANTINED_LEAK_DETECTED' but description does not explain what LLM should do next (retry, ask user, abort). No recovery path documented.