Loading ToolBench
Autonomous form autofill resolver that heuristically maps buyer profile to checkout web forms
Single tool with critical definition gaps. Tool name 'resolve_form_autofill' lacks action verb prefix (should be 'autofill_form' or 'match_form_fields'). Description is 60 chars, acceptable length but vague ('Heuristically maps' lacks specificity on what mapping means or when to use it). Input schema visible but parameters lack descriptions entirely, user_profile and form_fields have type/optional info but no guidance on expected structure, required fields, or format. No output schema documented. No error handling guidance. Tool definition inferred from code rather than explicit MCP registration with full schema.
Heuristically maps buyer profile to checkout web forms.
Tool name does not start with action verb. 'resolve_form_autofill' is passive/unclear; should be 'autofill_form' or 'match_form_fields' to signal intent.
Input parameters lack descriptions. 'user_profile' and 'form_fields' have type/optional metadata but no text explaining expected structure, required subfields (e.g., 'first_name', 'email'), or format constraints. LLMs cannot infer parameter meaning from names alone.
No output schema documented. Tool returns a complex object with 'autofill_id', 'field_assignments', 'status' but LLMs have no visibility into return structure, field types, or what to extract for downstream use.
No error handling or recovery guidance. Tool silently defaults to hardcoded test data if parameters are null; no indication of failure modes, validation errors, or what the LLM should do if autofill fails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | 2026-07-28+ | v2 |
Tool definition inferred from code rather than explicitly registered in MCP protocol. mcp_server.py hardcodes tool metadata in tools/list response without formal schema registration, capping confidence in definition completeness.