Loading ToolBench
MCP server for PromptEnhance - integrates with Claude Code to enhance developer prompts with relevant codebase context
PromptEnhance MCP server has a focused tool set (4 tools) targeting codebase context enhancement. Tool schemas are present with basic type information, but descriptions lack depth and guidance for LLM selection. Input parameters are typed but descriptions are minimal. Output schemas are not documented. Error handling is not evident in the source. The tool set follows verb_noun convention but lacks differentiation and composition clarity. Overall, the server is below median quality, it has structure but lacks the LLM-optimization and completeness expected for production use.
Enhance a developer prompt with relevant codebase context. Takes a natural language prompt and returns an enriched version with code snippets, file structure, conventions, and dependencies.
Get information about the indexed project including file count, languages, dependencies, and conventions.
Index or re-index the current project. Scans all code files, extracts structure, and creates embeddings for semantic search.
Search the codebase semantically for relevant code based on a query.
Output schemas not documented. Users cannot see what enhance_prompt, index_project, get_project_info, or search_codebase return. LLMs cannot plan downstream tool composition without knowing output fields.
Parameter descriptions are minimal or missing context. 'prompt' in enhance_prompt lacks guidance on format (natural language? code snippets?). 'query' in search_codebase lacks examples or format hints. Parameter descriptions should be 50-150 chars explaining what the parameter controls and acceptable formats.
Tool descriptions lack LLM-selection context. 'Enhance a developer prompt with relevant codebase context' for enhance_prompt does not explain WHEN to call it vs search_codebase, or what prerequisites exist (must index_project first?). Descriptions should answer: What? When? Why distinct from similar tools?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 49 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 49 | - | v1 |
No error handling or recovery guidance documented. If index_project fails, what should the LLM do? If search_codebase returns no results, is that a failure or normal? Missing error categorization (retryable vs user-fixable vs fatal) and recovery paths.
Tool composition unclear. Must users call index_project before search_codebase or enhance_prompt? No dependency hints in descriptions. Tools should state prerequisites: 'Call index_project first if this is the first run.'
Parameter defaults present (maxTokens=4000, includeGitContext=true, force=false) but rationale not explained. Why 4000 tokens? Is that suitable for all codebases? Why include git context by default? Defaults should be documented as reasonable common cases.
search_codebase 'topK' parameter (default 10) lacks context. Is 10 results always sufficient? What if user needs 100? Should topK have min/max bounds? Description should specify acceptable range.