Loading ToolBench
MCP server for CRM marketing campaign management, enabling creation of campaigns, targeting rules, email content, and execution schedules
Four tools with complete input schemas and descriptions, but significant quality gaps. All tools follow verb_noun naming (create_*). Descriptions are present (avg 85 chars) but lack actionable context, they state WHAT but not WHEN or WHY to use each tool. Parameters have types and descriptions, but lack constraints (enums, ranges, formats). No output schemas documented. Error handling absent. No guidance on tool sequencing or dependencies (e.g., create_campaign must precede create_targeting_rule). Security concerns: SQL injection risk in create_targeting_rule (raw sqlQuery parameter). No idempotency guidance for retry scenarios.
create new campaign for CRM marketing
create execution schedule for CRM campaign
create email content for CRM campaign
create targeting rule for CRM campaign using SQL query
SQL injection vulnerability in create_targeting_rule: sqlQuery parameter accepts raw SQL with no sanitization or validation. LLM could be tricked into passing malicious payloads.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required IDs (cpgnRid) for tool chaining.
Descriptions lack dependency hints and sequencing guidance. No indication that create_campaign must be called first, or that its cpgnRid output feeds into other tools.
Date/time format constraints (YYYYMMDD, HHMMSS) stated in descriptions but not enforced as regex patterns in schema. LLMs frequently misformat dates.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | <=2025-11-25 | v2 |
purposeCode parameter in create_campaign lists enum values in description text only; not declared as formal enum in schema. LLM may hallucinate invalid codes.