PromptShield sanitizer that applies safe, deterministic fixes to text based on detected prompt-injection threats such as invisible characters, markdown smuggling, and BOM artifacts.

Local-first, deterministic security scanner for Model Context Protocol (MCP) servers. Cross-tool toxic-flow analysis, Unicode-smuggling decode, prompt-injection & supply-chain detection, with an auditable 0–100 Trust Score.

The heart of the PromptShield ecosystem. A zero-dependency, isomorphic TypeScript engine for detecting invisible characters, BIDI overrides, and homoglyph attacks in AI prompts.

Two offline engines for AI tooling: secret detection (regex + Shannon entropy) and prompt-injection detection (Unicode smuggling, bidi overrides, homoglyphs). Zero runtime dependencies. The core of the Quell VSCode extension.

Sanitize untrusted content to prevent Indirect Prompt Injection (IPI) in RAG pipelines — strips invisible Unicode smuggling (zero-width, bidi, tag characters) and neutralizes injection markers. Zero dependencies, fully typed.

Deterministic LLM prompt defense scanner — 17 attack vectors + 12 Unicode smuggling categories + 6 output-scanner rules for MCP/agent attack surface. Pure regex, zero AI cost, < 5ms.

Secure, zero-dependency minimist replacement with prototype pollution protection, Unicode smuggling defense, NFKC normalization, deterministic alias groups, and a sandboxed unknown hook. Drop-in compatible.

Official Tiptap extension for PromptShield - Protect your editor from Prompt Injection and Trojan Source attacks in real-time.

Cupel — audit local des skills IA (Claude Code, Cursor, Codex). 14 règles de détection : prompt injection, ASCII smuggling, tool poisoning, exfiltration credentials, reverse shells, obfuscation hex. Zero network. Inspiré de la coupelle de l'essayeur d'or,

Framework-neutral opaque application sessions with secure H3 helpers

Six-tier memory system for the Graphorin framework: createMemory() facade with working / session / episodic / semantic / procedural / shared sub-modules, eleven memory tools wired into @graphorin/tools, hybrid vector + FTS5 search composed through Recipro

Better Auth plugin for Tauri desktop apps — handles OAuth via the system browser with deep-link callbacks, cookie bridging via URL, and secure token storage hooks.

Scan a repo for hidden prompt-injection payloads before your AI agent trusts it.

53j1sc

Unicode spoofing detection: mixed-script words, UTS #39 confusables, invisible characters, and combining-mark abuse. Zero runtime dependencies.

You vibe-coded your agents. Time for a vibe check. Linter and security scanner for Claude skills, agents, and AI workspaces.

A lightweight MCP security guard for Claude Code users and indie AI builders. Claude Codeユーザーのための、接続前MCPセキュリティ診断ツール。

Black-box & white-box security auditor for web applications with HackTricks techniques

Delegate safely without losing control — visible workers, evidence-first verdicts, and durable deliverables outside model context

Catches hidden and invisible-unicode instructions smuggled into AI coding-agent config and skill files (CLAUDE.md, AGENTS.md, .cursorrules, copilot-instructions.md, MCP, skills). Zero-config.