Loading ToolBench
Multi-module agent platform with blueprint engine, MCP servers, artifact store, prompt registry, and CLI tools for AWS-based agent orchestration
Server provides 10 tools with basic schemas and descriptions, but quality is inconsistent. Most tools have descriptions (8/10), but many are generic or lack actionable detail. Input schemas are present but incomplete: parameters lack type constraints (enums, ranges), and output schemas are not documented. Memory tools expose session_id as a parameter (security concern). Tool naming is reasonable (verb_noun pattern), but descriptions average ~60 chars, below the 194-char baseline for A+ tools. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint, destructiveHint). Overall falls into the 'fair' category with significant gaps in parameter validation, output documentation, and error guidance.
Send a query to a remote specialist agent and return the response via A2A protocol or direct invoke
Create and store an artifact using the claim-check pattern (S3 + DynamoDB)
Find tools by keyword matching against names and descriptions from the Gateway
Find tools relevant to a complex task using weighted scoring with name, description, and target context matching
Retrieve a stored artifact by ID
List all available tools from the Gateway
List all available artifacts
Memory tools expose session_id and actor_id as required parameters. These are internal identifiers that should be injected server-side via authentication context, not passed by the agent. Exposing them risks session hijacking and violates the secret-injection pattern.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., artifact_id from create_artifact) without knowing the response structure. This forces agents to guess and wastes tokens on parsing.
Parameters lack type constraints. 'type' in create_artifact has no enum; 'max_results' in find_tools has no min/max bounds; 'timeout_s' in poll_artifact has no range. Free-form strings and unbounded numbers invite hallucinated or absurd values.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Recall stored memories from the agent's memory system
Record and store memories in the agent's memory system
Poll for artifact completion with timeout
Descriptions are generic and lack actionable context. 'Recall stored memories' (memory_recall) does not explain when to use it vs memory_record, what format memories are stored in, or what happens if memory_id is not found. Average description length ~60 chars vs 194-char baseline.
No error handling or recovery guidance. Tools do not document what happens on failure (e.g., artifact not found, timeout exceeded, remote agent unreachable). Agents have no guidance on retry strategy or fallback actions.