### Benchmark Prompt Example for Glootie Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md This is an example of a detailed benchmark prompt for the Glootie system. It outlines specific instructions for continuous tracking, hypothesis testing, implementation strategies, code quality standards, and error handling. ```text ``` always use glootie ``` my preferred prompts look somethinig like: ```always continuously track and update caveats with the caveat tool, always hypothesize and test ideas in glootie and playwright execute first before implementing them, only implment when you're sure something will work, use the tooling to always eliminate double implementations, DRY code is mandatory, generalization is mandatory, architectural foresight is mandatory, immediately implement changes that should be made across the board, use the code insight tools in glootie to improve your ourput and immediately fix anything that goes against policy, a stitch in time saves nine. use ast-grep for codebase wide pattern base replaces etc whenever needed, dont make any comments always eliminate any coments you see immediatley, dont make any mocks simulations fallbacks or failovers, our code must only have one primary implementation per concern, keep specs in specs/ and track their progress, you are not finished till all specs are recorded, tracked, all the parts of the codebase they use noted, end to end verified, and caveats recorded, never report further problems and finish, if they can be solved track them troubleshoot and iterate until they're fixed, if any problems warnings issues or errors or unexpected behaviors are encountered track them immediately and treat the solving of all known errors as a hard barrier that prevents you from finishing, you may only finish when all known issues are fully end-to-end verified to be resolved, that means that as a final step you must do a full error discovery run to find further issues and check the entire project to make sure it works as expected, end to end, tracking any new issues found immediately and iterating on troubleshooting them till an entire run of all the program features are fully end to end tested and known to work, no exceptions. we want to apply frameworking foresight thoughout our work process so that generalizations, dryness, automated frameworking, convention over configuration and code minimization is always enforced, every letter counts. we want compehensive and easy to use globals for debugging in our code and often use them in our code executions as possible, to get absolute truth on whats going on when fixing problems, we should never ever guess when changing code, first principals only.``` ``` -------------------------------- ### Execute Multi-Language Code Examples Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md Examples of how to use the 'execute' tool for running code in various programming languages. This tool supports Node.js, Deno, Go, Rust, Python, C, C++, and Bash, with automatic runtime detection. ```shell # Example: JavaScript/TypeScript execution execute --lang js --code "console.log('Hello from Node.js!');" # Example: Go execution execute --lang go --code "package main; import \"fmt\"; func main() { fmt.Println(\"Hello from Go!\") }" # Example: Python execution execute --lang py --code "print('Hello from Python!')" # Example: Bash execution execute --lang sh --code "echo 'Hello from Bash!'" ``` -------------------------------- ### ast_tool examples for code analysis and manipulation Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md These JavaScript examples demonstrate the usage of the `ast_tool` for various code operations. They cover analyzing code structure, searching for specific patterns like React components, replacing `console.log` with a logger, and applying custom linting rules. ```javascript // Analyze code structure ast_tool(operation="analyze", path="./src", analysisType="detailed") // Find React components ast_tool(operation="search", pattern="const $NAME = ($PROPS) => { $BODY }") // Replace console.log with logger ast_tool(operation="replace", pattern="console.log($MSG)", replacement="logger.info($MSG)") // Lint for code quality issues ast_tool(operation="lint", path="./src", rules=[custom_rules]) ``` -------------------------------- ### Add mcp-glootie to Claude Code on Windows Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md This command adds the mcp-glootie MCP server to Claude Code on Windows. It uses `npx` to install and run the package and requires the `cmd /c` wrapper for `npx` commands. ```bash claude mcp add glootie -- cmd /c npx -y mcp-glootie@latest ``` -------------------------------- ### Configure GitHub Copilot MCP server for glootie Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md This JSON configuration is for GitHub Copilot's `mcpServers.json` file. It sets up the `glootie` server using `npx` to run `mcp-glootie@latest`, specifying it as a local type and listing the available tools. ```json { "mcpServers": { "glootie": { "command": "npx", "args": ["-y","mcp-glootie@latest"], "env": {}, "type": "local", "tools": [ "execute", "searchcode", "ast_tool", "caveat" ] } } } ``` -------------------------------- ### Caveat Management with Caveat Tool Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md Shows how to use the 'caveat' tool to manage technological caveats during development. This includes recording, viewing, and deleting limitations, constraints, and considerations, which is crucial for tracking project knowledge. ```shell caveat --record --message "API rate limiting may cause issues under heavy load." caveat --list caveat --delete --id 123 ``` -------------------------------- ### Unified AST Operations with ast_tool Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md Illustrates the use of the 'ast_tool' for unified Abstract Syntax Tree operations. This includes code analysis, pattern searching, safe code replacement, and linting, providing a comprehensive way to interact with code structure. ```shell ast_tool --action analyze --file "./src/main.js" --pattern "console.log(*)" ast_tool --action replace --file "./config.yaml" --find "old_setting" --replace "new_setting" ``` -------------------------------- ### Configure Cursor MCP server for glootie Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md This JSON configuration is for the Cursor IDE's `mcpServers.json` file. It defines the `glootie` server, specifying `npx` as the command to run the `mcp-glootie` package with specific arguments and enabling various tools like `execute` and `searchcode`. ```json { "mcpServers": { "glootie": { "command": "npx", "args": ["-y","mcp-glootie@latest"], "env": {}, "disabled": false, "autoApprove": [ "execute", "searchcode", "ast_tool", "caveat" ] } } } ``` -------------------------------- ### Run Test Runner Script with Node.js Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md The project includes a comprehensive performance testing suite executed via `test-runner.cjs`. This script runs baseline vs MCP-optimized tests simultaneously, monitors performance, and generates reports. It is executed using Node.js. ```bash node test-runner.cjs ``` -------------------------------- ### Execute JavaScript Code with Optimization Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md The execute tool allows for multi-language code execution with a 3-second threshold optimization and cross-tool status sharing. It supports various runtimes, including Node.js for JavaScript. Fast executions (< 3 seconds) return direct responses and status is automatically shared. ```javascript execute(code="console.log('Hello world')", runtime="nodejs") ``` -------------------------------- ### Manage Technological Caveats with Provided Text Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md The caveat tool is used to record, view, and delete technological caveats. This helps track system limitations, constraints, and important considerations. Caveats are stored locally and displayed during MCP initialization to inform future work. It supports recording by text. ```javascript caveat(action="record", text="This API endpoint has rate limiting of 100 requests per minute") ``` ```javascript caveat(action="view") ``` ```javascript caveat(action="delete", id="cav_1234567890123_abcde") ``` -------------------------------- ### Search Code with AI-Powered Vector Embeddings Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md Demonstrates the usage of the 'searchcode' tool for performing semantic code searches using AI-powered vector embeddings across multiple programming languages. This enables advanced code discovery and analysis. ```shell searchcode --query "function to find prime numbers" --language python ``` -------------------------------- ### Configure VSCode MCP server for glootie Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md This JSON configuration is for VSCode's MCP settings. It defines the `glootie` server using `npx` to execute `mcp-glootie@latest` with standard input/output communication (`stdio`). ```json { "servers": { "glootie": { "command": "npx", "args": ["-y","mcp-glootie@latest"], "env": {}, "type": "stdio" } }, "inputs": [] } ``` -------------------------------- ### Add mcp-glootie to Claude Code on macOS/Linux Source: https://github.com/anentrypoint/mcp-glootie/blob/main/README.md These commands add the mcp-glootie MCP server to Claude Code on macOS and Linux. The recommended method uses `npx` for the latest version, while the local development option allows direct execution of a local script. ```bash # Using npx (recommended - always gets latest version) claude mcp add glootie npx -s local -- -y mcp-glootie@latest # For local development (replace /path/to with actual path) claude mcp add glootie -- node /path/to/mcp-glootie/src/index.js ``` === COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.