Loading ToolBench
Two MCP servers for Week-04 agentic incident command: a minimal tool harness warm-up and a full incident command agent with memory store, resource management, and structured telemetry.
This server has critical definition gaps. 3 of 8 tools (add_evidence, append_delta, write_plan) have empty or missing input schemas, automatic 0 for schema. 5 tools lack parameter descriptions entirely. Tool names are inconsistent (append_memory_delta vs append_delta; summarize_incident vs create_incident). Descriptions are present but generic (10-50 chars). No output schemas documented. No error handling guidance. This is a proof-of-concept harness, not production-ready.
Add evidence to an incident record.
Append a delta to the memory store.
Append a structured loop-outcome delta to memory://deltas/recent.
Create or update an incident record in memory.
Retrieve runbook snippets by keyword.
Return canned diagnostics for a host.
Summarize the incident with evidence.
Write OPAL plan into memory://plans/current
Three tools (write_plan, append_memory_delta, add_evidence, append_delta) have empty or missing input schemas.
add_evidence and append_delta have no descriptions at all.
Tool names are inconsistent and vague: 'append_memory_delta' vs 'append_delta' (same intent?), 'write_plan' (to where?), 'create_incident' vs 'summarize_incident' (both modify state but names don't clarify). Ambiguous names cause LLM confusion.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | <=2025-11-25 | v2 |
No error handling guidance. Tools return success/failure but do not tell LLM what to do on error (retry? ask user? fatal?). No recovery paths documented.