Loading ToolBench
LLM-powered cross-platform chat topic guardian with multi-adapter support, detection rules, user profiling, and MCP integration
ChatGuardian exposes 17 tools with complete input schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention (get_, create_, update_, delete_) appropriately. Descriptions are present but often generic (e.g., 'Retrieve detection results with optional filtering' lacks context on when to use vs. get_detection_rules). Parameter descriptions exist but lack constraints, no enums for provider choices, no ranges for numeric thresholds. Output schemas are not documented in the tool definitions. Error handling is absent from descriptions. Security-sensitive operations (delete_detection_rule, update_llm_config) lack permission gates or confirmation patterns. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels assigned in metadata.
Send a message to the agent and get a response
Create a new agent chat session
Create a new detection rule with specified parameters
Delete a detection rule by ID
Get status of all configured message adapters
Retrieve agent chat sessions
Retrieve chat message history with optional filtering
No output schemas documented. Tool definitions specify inputs but omit return types, forcing LLMs to guess what fields are available for downstream chaining.
Enum constraints missing for constrained parameters. 'provider' in update_llm_config accepts 'openai, ollama, google, anthropic' but is defined as free-form string; 'notifier_type' in test_notification lacks enum; 'log_level' in update_system_settings is free-form despite fixed options.
Destructive operations lack confirmation or dry-run patterns. delete_detection_rule and update_system_settings (which modifies queue/history limits) have no confirmation step, inviting accidental data loss.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Retrieve detection results with optional filtering
Retrieve all detection rules from the system
Retrieve current LLM configuration
Retrieve notification configuration
Retrieve system configuration settings
Retrieve user behavior profile and memory data
Send a test notification to verify configuration
Update an existing detection rule
Update LLM configuration settings
Update system configuration settings
Descriptions lack actionable context. Generic phrases like 'Retrieve detection results with optional filtering' do not explain when to call this vs. get_detection_rules, what filtering options exist, or what the results contain.
No error handling guidance in tool descriptions. Tools do not document what errors are retryable, what user-fixable errors look like, or how to recover from failures.