Rigour Labs | AI Agent Governance — DLP, Quality Gates & Memory Control

# Rigour Labs > AI Agent Governance. One command. Every agent. Block credential leaks, enforce quality gates, control what agents remember. ## What is Rigour? Rigour is the governance layer for AI coding agents. It provides three layers of protection: Input DLP (blocks credential leaks with 29 patterns), Quality Gates (deterministic PASS/FAIL on every file write), and Memory Governance (controls what agents remember). Works with Claude, Cursor, Cline, Windsurf, and Copilot. 100% local, zero telemetry. ## Core Problem AI agents leak credentials, write ungoverned code, and persist data in uncontrolled memory files. Without governance, every AI coding agent is a compliance risk. Rigour adds the missing control layer. ## Three Layers of Protection ### 1. Input DLP (Data Loss Prevention) Scans every agent input for credentials before they reach the AI model. - 29 credential patterns: AWS keys, GCP service accounts, Azure tokens, API keys (OpenAI, Anthropic, GitHub, Stripe, SendGrid), database URLs, JWTs, private keys - Shannon entropy detection catches base64/hex encoded secrets - Unicode normalization defeats zero-width character bypass attempts - OWASP LLM Top 10: 10/10 coverage ### 2. Quality Gates Real-time hooks fire on every file write across all AI agents. - Deterministic PASS/FAIL: file size, complexity, security patterns, hallucinated imports, AI drift - Deep LLM analysis across 40+ code quality categories - Supports: TypeScript, JavaScript, Python, Go, Rust, Java ### 3. Memory & Skills Governance Controls what agents persist and where. - Blocks writes to native agent memory: .cursorrules, CLAUDE.md, .clinerules - Blocks writes to agent skills: .claude/skills/, .cursor/rules/, .windsurf/rules/ - DLP scanning on recall — blocks tainted stored credentials - Configurable via rigour.yml ## Install ```bash npx @rigour-labs/cli init # or brew tap rigour-labs/tap && brew install rigour ``` ## Products ### Rigour CLI Core governance engine. Runs locally w