Loading ToolBench
MCP Server pour générer automatiquement des Pull Requests parfaites
This server has moderate definition quality with significant structural issues. Tool descriptions are present but often confusing, many repeat warnings about returning 'prompts' rather than final outputs, which violates clarity principles. Parameter schemas are defined with types and descriptions, but composition is poor: tools that should be unified (generate_pr_title, generate_pr_description, review) return analysis prompts instead of actionable results, forcing the LLM to do work that belongs in the tool. All 8 tools have descriptions and schemas, but the conflation of 'prompt generation' with 'tool output' undermines usability. Error handling is minimal, no guidance on recovery or categorization. The average tool-level score is 52 across naming (fair), descriptions (confusing), and schemas (present but incomplete).
Analyzes the differences between the current Git branch and a base branch. Automatically detects the main branch (dev/main/master) and current Git repository from the working directory. Use this when the user asks to analyze a branch, check changes, or understand what's in the current branch.
Creates a Pull Request on GitHub WITHOUT AI review. **RECOMMENDED WORKFLOW**: (1) Call 'generate_pr_title' → get prompt, (2) YOU analyze and generate title, (3) Call 'generate_pr_description' → get prompt, (4) YOU analyze and generate description, (5) Call THIS tool with 'title' and 'description' parameters. If not provided, simple versions are auto-generated from commits/templates. NOTE: This tool does NOT include AI code review. For complete PR WITH AI review, use 'create_pr_complete' instead. This tool: analyzes branch, creates/updates PR, adds reviewers automatically. Requires GITHUB_TOKEN. Use when user asks to create a simple PR without code review.
Creates a Pull Request on GitHub WITH integrated AI code review. **RECOMMENDED WORKFLOW FOR AI ASSISTANTS**: (1) Call 'generate_pr_title' tool to get an AI prompt, (2) YOU (Claude) analyze and generate an intelligent title, (3) Call 'generate_pr_description' tool to get an AI prompt, (4) YOU (Claude) analyze and generate an intelligent description, (5) Call 'review' tool to get comprehensive context/diff prompt, (6) YOU (Claude) analyze and write a code review, (7) Call this tool with 'title', 'description', and 'aiReviewText' parameters. This tool performs the complete PR creation workflow: analyzes changes and gathers project context, uses provided title/description or generates simple versions, prepares the AI review section (if aiReviewText provided), adds relevant GIF at the end (after review), creates/updates the PR on GitHub, assigns appropriate reviewers automatically. Requires GITHUB_TOKEN. Use when user asks to create a complete PR with code review included.
Three tools (generate_pr_title, generate_pr_description, review) are described as returning 'ANALYSIS PROMPTS' for the LLM to process, not final outputs. This violates the tool-as-action pattern, tools should return actionable data, not prompt templates for the agent to parse and reimplement. This shifts responsibility incorrectly and creates confusion about tool semantics.
Descriptions for generate_pr_title, generate_pr_description, and review are confusing and lengthy (>300 chars), with heavy emphasis on what the LLM should do rather than what the tool does. Descriptions should state tool behavior, not coach the agent. E.g., generate_pr_title description is 451 chars and reads like a prompt template ('YOUR JOB: Analyze...') rather than a tool contract.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 39 | - | v1 |
Generates both a conventional commit-style PR title and a complete Pull Request description in one operation. Combines the functionality of generate_pr_title and generate_pr_description. Automatically analyzes the current Git branch from the working directory. Use this when the user wants both title and description generated together.
⚠️ CRITICAL: This tool returns an ANALYSIS PROMPT for YOU (Claude) to generate a description, NOT a final description. OUTPUT: You receive comprehensive context (project type, commits, files, diff sample) formatted as a detailed prompt with specific instructions. YOUR JOB: Analyze that context and generate an intelligent, well-structured PR description. The prompt includes format requirements (What/Why/How/Impact sections), language preferences (FR/EN), and style guidelines. YOU must read it, understand the changes, and write a clear description ready for GitHub. DO NOT return the prompt - generate YOUR description. Use when creating PR or when user asks for PR description. The generated description should then be passed to create_pr or create_pr_complete tools.
⚠️ CRITICAL: This tool returns an ANALYSIS PROMPT for YOU (Claude) to generate a title, NOT a final title. OUTPUT: You receive comprehensive context about the code changes (commits, files, diff patterns) formatted as a prompt. YOUR JOB: Analyze that context and generate an intelligent conventional commit-style PR title (e.g., 'feat(auth): add OAuth support'). The prompt provides all the information you need to understand what changed. YOU must read it and create a concise, descriptive title that accurately reflects the changes. DO NOT return the prompt - generate YOUR title. Use when creating PR or when user asks for a PR title. The generated title should then be passed to create_pr or create_pr_complete tools.
⚠️ CRITICAL: This tool ONLY returns an ANALYSIS PROMPT for YOU (Claude), NOT a final review. OUTPUT: You receive comprehensive context (project type, diff, commits, structure) formatted as a detailed prompt with instructions. YOUR JOB: Analyze that context and generate the actual code review following the exact format specified in the prompt. The prompt includes everything you need: project context, full diff, testing framework, architecture details. YOU must read it, analyze the code changes, and write a concise review (max 10-15 lines) with sections: Summary, Critical Issues, Key Suggestions, Decision (APPROVE/REQUEST_CHANGES). DO NOT return the prompt itself - generate YOUR review. Works for ANY language/framework because context is provided. Use when user asks to review code or when creating PR with review.
Suggests code reviewers based on Git contribution history of modified files in the current branch. Automatically analyzes the working directory's Git repository. Use this when the user asks who should review the PR or to suggest reviewers.
No input validation, error categorization, or recovery guidance in visible code. If GITHUB_TOKEN is missing, create_pr and create_pr_complete throw a generic error with no suggestion for next steps. No distinction between retryable and permanent failures.
create_pr and create_pr_complete accept githubToken as a parameter, exposing credentials. Best practice is server-side secret injection via environment variables only. Accepting tokens as parameters risks logging and prompt leakage.
Three pairs of tools (generate_pr_title + generate_pr_complete, generate_pr_description + generate_pr_complete, review + create_pr_complete) have overlapping responsibilities and no clear composition guidance. An LLM may choose the wrong variant. Consolidate or clearly document when to use each.
Parameters with enums (template, language) lack descriptions explaining the difference between options. 'standard', 'detailed', 'minimal' templates are mentioned but not explained. What is the use case for each? How do they differ?
Output schemas are not explicitly documented in tool definitions. AnalysisResult, CreatePRResult, and CreatePRCompleteResult types are referenced but their fields are not shown in the tool schema documentation visible to the agent.