Loading ToolBench
An MCP server that generates video scripts based on topics and keywords using AI agents. It supports both short-form (reel) and long-form video script generation with multiple specialized agents for different aspects like tone, content, intro hooks, and formatting.
Script Generator Server has a single tool 'script_generate' with incomplete and minimal quality. The tool has a valid schema with typed parameters and descriptions, but lacks depth in its documentation. The description is trivial (11 words), providing no guidance on when to use the tool, what it does internally, or what the return value structure is. The schema declares topic and keywords as required strings, but provides no constraints, ranges, or format guidance. Output schema is completely undocumented, the tool returns a string via `await script_using_topic(task)`, but there's no specification of what that string contains, its structure, or how an LLM should parse it. Tool naming is acceptable but generic ('script_generate' is action-driven), but naming alone cannot compensate for missing descriptions and output documentation. The server includes prompts and resources, but these are minimally populated and serve no clear purpose in the context of the script generation task.
Provide topic and keyword to generate Script
Missing output schema documentation. Tool returns a string via `script_using_topic(task)`, but there is no specification of the return structure, content format, or how an LLM should interpret or use the result.
Tool description is trivial (11 words: 'Provide topic and keyword to generate Script'). It does not explain WHAT the script output is (e.g., video script? marketing copy? dialogue?), WHEN to use it, or what the LLM should expect. LLMs cannot select or compose tools based on this.
Parameters lack format guidance and constraints. 'topic' and 'keywords' are declared as strings but have no length limits, format patterns, or examples. An LLM cannot distinguish between a 5-char topic and a 5000-char topic, or determine if keywords should be comma-separated, space-separated, or newline-separated.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 32 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 33 | - | v1 |
Tool behavior is opaque. The description mentions using keywords 'as reference' but the implementation concatenates keywords into a verbose template that includes confusing instructions (Time, Objective, Audience, Gender, Tone, Speakers). The function's actual purpose, what kind of script, what audience, what length, is never clarified.
No error handling or recovery guidance. If `script_using_topic()` fails (timeout, API error, authentication issue), the tool offers no recovery path, retryability classification, or user-fixable error message. LLMs will not know whether to retry, ask the user, or give up.
Prompts and resources are present but unused. The server declares 'summarize-notes' prompt and 'note://' resources, but these are not connected to script generation. They appear to be boilerplate from a template and add noise without value.