#prompt-injection

4 resources and 3 guides tagged #prompt-injection on ChangeGamer.

Agentic Security Checklist · Guide
Cross-vendor, threat-surface-organized security checklist for building and operating AI agents — synthesizing OWASP, NIST, Anthropic, OpenAI, Google SAIF, and MITRE ATLAS.
Guardrails and Safety Filters for Agents · Guide
Runtime input/output/action controls that enforce policy independently of the model — tooling landscape, techniques, and layering guidance.
Agentic AI Browsers: Comet, Atlas, and the Prompt-Injection Attack Surface · Guide
What agentic browsers (Perplexity Comet, the now-sunsetting ChatGPT Atlas, Microsoft Edge Copilot Mode, Opera Neon) are, how they differ from developer-facing computer-use APIs, and the documented prompt-injection attacks — CometJacking, indirect injection, hidden-text/screenshot instructions — that target the whole product category.
Prompt Injection Design Patterns: Architectural Defenses for Agents · Guide
Six named architectural patterns — Action-Selector, Plan-Then-Execute, LLM Map-Reduce, Dual LLM, Code-Then-Execute, Context-Minimization — plus Google DeepMind's CaMeL, that structurally constrain what an agent can do with untrusted data instead of just filtering it.

Guides

Defending MCP Clients Against Tool Description and Output Injection · guide
Two distinct MCP injection surfaces — a tool description at connect-time and a tool's return value at call-time — and the client-side architectural patterns (Dual LLM, Action-Selector, Context-Minimization) that contain each one.
How to Secure AI Agents in Production · pillar guide
Credential hygiene, prompt-injection defense in depth, sandboxing choices for code execution, supply-chain provenance, least privilege, audit trails, incident response, and rate/abuse controls — eight operator-side defenses against an adversarial actor or a compromised dependency, not against ordinary load or failure.
How to Defend an AI Agent Against Prompt Injection · guide
A decision framework for matching Action-Selector, Plan-Then-Execute, Dual LLM, and the other named architectural patterns to a task's actual blast radius, including when to compose two patterns together and when none of them is worth the overhead.