sleuthre-mcp exposes 17 reverse-engineering tools with consistent naming (verb_noun pattern: get_*, open_*, rename_*, add_*, save_*, load_*, run_*) and complete input schemas. However, descriptions are terse (10-50 chars, well below the 194-char baseline), lack context on WHEN to use each tool, and omit dependency hints. Parameters are typed but lack detailed constraints (e.g., address format, count bounds). Output schemas are not documented. Error handling is absent from descriptions. Tools are well-scoped (single responsibility), but the interface prioritizes reverse-engineering domain knowledge over LLM-friendly guidance.

• Expand all tool descriptions to 100-200 chars. Include WHAT the tool does, WHEN to use it (e.g., 'Use get_hlil for human-readable decompiled code; use get_mlil for SSA-form intermediate representation'), and any prerequisites (e.g., 'Requires open_binary to be called first').
• Document output schemas for all tools. For list tools (get_functions, get_strings, get_imports, get_exports), specify the structure of each item, include pagination guidance (limit, offset, total_count), and cap results at 50 items with a note in the description.
• Add parameter constraints to descriptions. E.g., 'address: hex string in format 0xHHHH, must be within binary bounds' and 'count: integer 1-256, default 16'.
• Add error handling guidance to descriptions. E.g., 'If the binary format is unsupported, returns error with list of supported formats (ELF, PE, Mach-O).' This helps LLMs self-correct.
• Document state dependencies explicitly. Add a note to all analysis tools: 'Requires open_binary to be called first with a valid binary path. Returns error if no binary is loaded.' Consider a get_current_binary tool to let LLMs verify state.