Three tools with clear semantic purpose (search, retrieval, RAG-powered QA). All have descriptions and input schemas present. Naming follows verb_noun pattern (search_docs, get_doc_section, ask_berachain). However, descriptions lack depth on WHEN to use each tool vs. alternatives, parameter descriptions are minimal (topK, maxContextChunks lack range/constraint details), and output schemas are not documented. No error handling guidance visible. Schemas are present but sparse, missing details on what fields are returned and how to chain results across tools.
Ask any question about Berachain development and get an AI-powered answer. This wizard tool uses RAG to search documentation and generates comprehensive, contextual answers using an LLM. Perfect for getting help with Berachain concepts, code examples, integration guides, and more.
Retrieve a specific documentation section by file path. Useful when you know the exact location of the documentation you need.
Search Berachain documentation and guides using semantic search. Returns the most relevant document chunks with their metadata including file paths and section titles.
Output schemas not documented. Tools return results but LLM cannot predict field structure (e.g., does search_docs return 'chunks' or 'results'? Does each item have 'filePath', 'section', 'content'?). Forces LLM to guess and risks parsing failures.
Parameter descriptions lack constraints. 'topK' and 'maxContextChunks' have defaults but no min/max bounds stated. LLM could pass topK=1000 or negative values. Descriptions should specify '1 - 20' or similar.
No error handling guidance. If a search returns zero results, or if the LLM asks an unanswerable question, what does the tool return? No recovery hints for the agent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Tool selection guidance incomplete. Server instructions say 'ask_berachain is the primary tool' but do not explain when search_docs is preferable (e.g., 'use search_docs if you want raw doc chunks; use ask_berachain for synthesis'). LLM may over-rely on one tool.
get_doc_section 'repo' enum includes 'local' but no description of what 'local' means or when to use it. Ambiguous for LLM.