sim-xia/blind-auditor

Overview

Blind-Auditor is an MCP tool that enforces a mandatory self-audit step before a model can output code. It intercepts the model's response, runs a structured audit, and only releases the code if the audit passes.

Best for

Best for
Developers who want a minimal guardrail to prevent obviously flawed model output from reaching production

Use cases

• Enforcing code quality checks in AI-assisted development workflows
• Adding a mandatory review gate before model-generated code is accepted
• Integrating self-audit into MCP-based coding agents

How to use

Install

uv sync

Tools exposed

• submit_draft
• submit_audit_result
• reset_session

Tested with

Cursor, Windsurf

Example client config

[object Object]

Notes

Blind-Auditor is an MCP tool that enforces a mandatory self-audit step before a model can output code. It intercepts the model’s response, runs a structured audit, and only releases the code if the audit passes.

10 stars on GitHub. Last updated 2025-12-10. Licensed MIT.

Use cases

• Enforcing code quality checks in AI-assisted development workflows
• Adding a mandatory review gate before model-generated code is accepted
• Integrating self-audit into MCP-based coding agents

Pros

• Simple mechanism to catch obvious errors before code is used
• Low overhead as a lightweight MCP tool
• Open source with Python codebase for easy customization

Cons

• Only 10 GitHub stars indicates very early stage or limited adoption
• Mandatory audit may slow down rapid prototyping workflows
• Effectiveness depends entirely on the quality of the audit logic

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.