Official TypeScript SDK for Lakera Red — run adversarial scans against your AI agents and chatbots.

OpenClaw plugin: screen tool calls with Lakera Guard and block if flagged

Defense adapters for prompt injection detection (Rebuff, Lakera, LLM Guard, Garak, Moderation APIs)

Pure-Rust prompt-injection detector with 1.5MB embedded MLP classifier. 98.40% accuracy, p50 14ms CPU inference, prebuilt napi addons for Linux/macOS/Windows. Apache-2.0/MIT alternative to Rebuff and Lakera Guard.

No description provided.

Deterministic conscience layer for AI agent output. Catches manipulative copy, false urgency, and dark patterns — the content-safety failure mode missed by input-attack guardrails.

CoFounder adapters package

Block AI assistant config leaks and secrets before npm publish

Weave Protocol browser-agent security — enforces WARD.md policies and detects indirect prompt injection (IPI) for Playwright, Puppeteer, and Stagehand-based AI agents

MCP server that redacts PII/PHI/secrets from outbound prompts before they leave your AI agent. Deterministic, local, auditable — built for CISOs preparing for EU AI Act Article 26 (Aug 2, 2026) and ISO 42001.

Content-aware provenance layer for Claude Agent SDK. Detects dangerous content at every tool I/O boundary, propagates classification across sub-agent dispatch, and enforces fail-closed policy.

On-device prompt injection detection for LLMs — patent-protected (FR2606828), zero data leaves your tenant. The first patent-backed open-source AI security SDK.