MCP Server for human-in-the-loop approval gate and execution interrupt control. Halts agent execution when actions exceed autonomy thresholds, creates pending review tickets with unique approval tokens, enforces role-based approvals, and emits resume tokens upon verified human authorization.
Single tool 'create_interrupt' has a verb-based name and basic schema, but critical gaps undermine quality. Description is 108 chars (acceptable range) but lacks WHEN to use it, prerequisites, or state-modification clarity. Input schema present but parameters lack descriptions, 'action' is typed as object with no guidance on structure; 'risk_level' has no enum constraint despite being a classification field. No output schema documented. No error handling guidance. Tool appears to be a write operation (WRITE risk) but description doesn't explicitly state it modifies state or has irreversible consequences. Missing idempotency guidance and confirmation pattern for high-risk operations.
Trigger a human-in-the-loop interrupt gate for high-risk agent operations.
Parameters lack descriptions. 'action' is typed as object but has no guidance on expected structure, keys, or format. 'risk_level' has no enum constraint despite being a classification field (HIGH, CRITICAL mentioned in description but not in schema).
No output schema documented. Callers cannot predict response structure (interrupt_id, status, approval_token, risk_level fields are visible in code but not declared in tool definition).
Description does not state that this tool modifies state (creates pending interrupt tickets) or has irreversible consequences. LLMs cannot determine if the call is safe to retry.
No error handling guidance. If thread_id is invalid, action_payload is malformed, or interrupt creation fails, the tool returns no recovery hints. LLM has no guidance on what to do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
High-risk operation (WRITE, CRITICAL risk level) lacks confirmation or dry-run pattern. Agents can trigger interrupts without a safety gate, risking accidental halts of legitimate operations.