Attack walkthroughs, integration examples and tests for your own app
Screen approved files and exact tool results with Claude Code hooks. Deny failed checks, keep tool permissions and test hook startup and timeout behavior.
Read OpenClaw’s sandbox defaults and three distinct research disclosures. Add a context gate before inference, then restrict tools and test authorized actions.
A generated app checks the shape of an input and trusts the words inside it. See the three inputs that carry instructions, and one ten-line check for them.
Hermes Agent reads web pages, MCP tool results, files and chat messages. See the four injection paths, three tested payloads, and where the check goes.
Grok on X, Grok Bot and your own Grok API app all read text you did not write. See the 3 injection paths, a real payload and verdict, and where to check.
Trace EchoLeak and a Supabase MCP research demonstration. Screen exact retrieved context, reject unavailable checks and restrict agent data access and actions.
Secure MCP agents by reviewing tool metadata, screening consumed context and enforcing exact-action permissions. Add strict error handling to your tool loop.
Screen the exact context before your LangChain Python model call. Use callbacks with explicit limits, strict verdict checks and separate tool permissions.
See how white text in an invoice reaches a PDF extractor. Inspect the archived example, screen assembled model context and test authorized application outcomes.
Chipotlai Max wrapped Chipotle’s Pepper support bot as a coding agent. See the project record, topic-policy gaps and controls for your own support bot.
Six tools compared on where each one runs, what it costs, and which projects are still maintained. Two of the six repositories are archived.
Screen Node.js model context before an OpenAI call. Use explicit Express request shapes, trusted client IPs, strict verdict checks and a pre-stream gate.
Compare Azure Prompt Shields setup, record billing and verdicts with a hosted alternative. Check request limits, external dependencies and fail-closed handling.
Add prompt screening before inference in Node.js or Python. Reject invalid fields, HTTP errors and malformed verdicts; forward only the exact accepted message.
Add checks before LangChain inference in Python, TypeScript and LangGraph. Reject failed verdicts, screen exact context and enforce tool permissions in code.
Review Claude MCP tool definitions, and screen requests and results before inference. Reject unavailable checks and authorize exact tool actions before execution.
Screen retrieved documents, web pages and email before inference. Include source labels, reject unavailable checks and keep tool permissions in your app.
Protect Next.js AI routes with a server-owned context gate. Check messages, retrieval and tools before inference, then open streams after a valid verdict.
Read OWASP’s LLM01 definition and seven mitigations. See where input screening fits, plus permission checks, output validation and application attack tests.
Learn how system prompt extraction attempts expose AI instructions. Keep secrets outside model context, screen incoming attacks and test authorized data access.
Screen retrieved chunks before the model reads them. Keep the four-layer RAG security model, validate the complete context, and stop safely when checks are unavailable.
LiteLLM 1.82.7 and 1.82.8 carried credential-stealing code. Trace the March 2026 incident, inspect offline and separate package security from prompt checks.
Compare prompt-security API formats and verdicts. Use an application adapter for provider changes, explicit error handling and tested migration behavior.
Read prompt injection payloads and detector verdicts from the dated playground dataset. Test these examples against your own app alongside ordinary requests.
Learn what prompt injection is, how direct and indirect attacks enter AI apps, and how input screening, data permissions and tool controls reduce the risk.
Explore the ten OWASP risks for LLM apps, each with a named case and practical control. Learn how input screening fits alongside other application defenses.
Learn how jailbreaking relates to prompt injection. Compare payloads, screen instruction attacks before inference, and enforce application data and tool limits.
Learn how untrusted documents, tool results and messages can redirect AI agents. Use input screening, permissions and exact-action approval before tools run.
Compare SQL value binding with prompt injection controls. Screen untrusted model context, enforce data and tool permissions, and test application outcomes.
Test the app through its real inputs, documents and tools. Define forbidden outcomes, preserve ordinary tasks, and make unsafe behavior fail CI.
Check the text, private data and tools your side project gives its AI. Choose prompt injection controls from its exposure, with a free input-screening plan.
A free Chrome extension that flags hidden text injection on web pages in real time, so you can inspect suspicious page text before sharing it with an assistant.
Everything you need to know about OpenClaw AI - the open-source personal AI assistant. Covers features, Moltbook integration, agent-to-agent security risks, and essential do's and don'ts for safe usage.
Explore 16 ways text can be concealed or obfuscated. Check what your AI extractor actually reads, then screen exact page context and keep action permissions.
Compare a poem in the browser with the server HTML your assistant may retrieve. Understand extraction differences and screen exact model context before use.
Map untrusted inputs, screen text before inference, authorize tools in your app, and test failure handling. A practical guide to layered protection.
Compare Lakera Guard with SafePrompt on pricing, deployment and verdict mapping. Review current plans and data handling, then test an application migration.
Learn why patterns miss reworded prompt injections. Read the dated Vigil mutation experiment and test semantic screening with attacks and ordinary traffic.
Choose an enforceable GPT protection path. Understand Action limits, check messages in a server you control, and authorize access to private records.
Keep your chatbot on its authorized task. Screen assembled context, enforce pricing and record permissions, then test attacks alongside ordinary requests.
Trace hidden email instructions from contact form to AI summary. Screen the submitted context, escape HTML, and authorize any resulting replies or tool actions.