Prevent LLM code generation hallucinations with a seven-layer defense framework.
npx skills add https://github.com/majiayu000/vibeguard --skill vibeguard
Please help me install this Agent Skill. Skill: vibeguard Source: https://github.com/majiayu000/vibeguard/tree/main/skills/vibeguard Command: npx skills add https://github.com/majiayu000/vibeguard --skill vibeguard
This Skill addresses the common failure modes in AI-generated code, such as inventing non-existent APIs, duplicating existing functionality, hardcoding data, and over-engineering solutions, by providing a systematic framework to prevent these issues.
/vibeguard:preflight to explore the project and generate constraints, ensuring the AI's output aligns with existing patterns and requirements.Run /vibeguard:preflight to explore the project and generate constraints before starting a new task.
High-intent search queries and answers about installing and using this skill.
A seven-layer defense architecture prevents AI code generation from inventing APIs and hardcoding data by injecting constraints, using real-time hooks for interception, and deploying specialized agents for review, testing, and fixing.
Adding guardrails for LLM code generation involves implementing a framework with real-time interception hooks and on-demand scanning tools to block dangerous commands and enforce coding constraints before execution.
To run a preflight check for project constraints before AI generates code, execute the `/vibeguard:preflight` command to explore the project structure and generate constraints ensuring output aligns with existing patterns.
Yes, this AI code review framework operates without external dependencies, utilizing specialized agents and MCP tools to perform on-demand scanning, real-time interception, and rule injection independently.
Real-time interception stops dangerous commands during AI-assisted development by utilizing hooks for immediate blocking, ensuring dangerous LLM code generation outputs are caught before execution.
anderson-0Generate project context documentation with implementation rules and technical conventions.
outfitter-devRun automated Navigator CLI tests with bun and navigator-server.
soy-rafaBlock malicious tool calls and scan installed skills for vulnerabilities.
LeoWSY-hashblueScan source code for vulnerabilities, secrets, and dependency CVEs.