A Docusaurus documentation site for the Jinshuju (金数据) open platform, featuring API documentation, webhooks, URL parameter passing, and form embedding. Includes client-side API debugging panel and URL parameter generator tools.
Both tools have adequate descriptions (60-80 chars) and clear parameter schemas with types. However, tool names violate verb-first conventions (should be 'generate_form_field_url' and 'generate_global_field_url'), and critical security issues dominate: sign_secret is exposed as a parameter despite claims of local-only computation. Parameter descriptions are present but lack format/constraint details. No output schema is documented. Error handling is absent. The tools appear to be browser-based generators (not server-side MCP tools), which raises fundamental architectural concerns about MCP suitability.
在线生成带签名的表单链接 (Generate form links with HMAC-SHA256 signature for field parameters)
在线生成带 JWT 的表单链接 (Generate form links with JWT (HS256) for global field parameters)
Tool names do not start with action verbs. 'form_field_url_params' and 'global_field_url_params' are noun-based; should be 'generate_form_field_url' or 'create_form_link_with_signature'. LLMs rely on verb-first naming to infer intent.
sign_secret exposed as a tool parameter despite documentation claiming 'local-only computation'. Credentials must never appear in tool parameters, they are logged in traces and prompt history. Use server-side secret injection via environment variables or vault.
No output schema documented. LLMs cannot plan downstream actions or extract the generated URL without knowing the response structure. Document what fields are returned (e.g., url, signature, payload).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
Parameter descriptions lack format constraints. 'form_token' should specify '6-character alphanumeric code from /f/ URL'. 'pairs' array items lack guidance on valid key formats (API CODE syntax). Vague descriptions invite invalid input.
No error handling or recovery guidance. If form_token is invalid or sign_secret is wrong, what error is returned? How should the LLM respond? Agents need actionable error messages, not silent failures.