Visa prepaid, eSIM, gift-card playbooks. No money. Read-only docs MCP: no secrets, no spend.
Critical failure across all dimensions. The server exposes only 2 tools (get_agent_facts, get_agent_use_cases) defined in scripts/ci-stdio.mjs. No tool schemas, descriptions, or parameter definitions are visible in any source file provided. The ci-stdio.mjs script only performs protocol initialization and lists tool names, it does not define the tools themselves. Tool definitions must exist in the @imba_wallet/agent-mcp-docs package (version 0.1.5), which is not provided in this evaluation. Based on the evidence available (tool names only), both tools cannot be scored above 50 per the HARD SCORING RULES. The server is read-only (no side effects), but without access to the actual implementation package, parameter schemas, descriptions, and output formats cannot be verified.
Tool definitions not visible in source. ci-stdio.mjs only lists tool names; actual schema, description, and parameter definitions reside in the unprovided @imba_wallet/agent-mcp-docs package. Cannot verify descriptions, parameter schemas, or output formats.
No tool descriptions visible. Both tools lack documented descriptions in the provided source.
No input schemas visible. Neither tool exposes parameter definitions.
Inferred tool existence. Tool names 'get_agent_facts' and 'get_agent_use_cases' appear only in test assertions within ci-stdio.mjs (lines checking if names are in listed tools). Combined with missing schema/description, actual overall per-tool score is 13.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |
No output schema documented. LLMs cannot plan downstream operations or extract required data without knowing response structure.
Vague/generic naming. 'get_agent_facts' and 'get_agent_use_cases' lack specificity about what facts/use-cases are returned or their domain (Visa, eSIM, gift cards). Names should indicate scope: e.g., 'get_visa_prepaid_facts', 'get_esim_use_cases'. Baseline: 90% of A+ tools start with action verb + clear noun.
No error handling guidance. No error responses documented. LLMs cannot know when to retry, ask the user, or escalate.