Results for “agent-security”

Expert en sécurité des agents IA (prompt injection, jailbreak prevention, tool misuse, output filtering)

The English summary is being prepared. Showing the original description.

Use when reviewing or writing LLM, RAG, MCP, tool, or agent code for OWASP-aligned security issues; triggered by "owasp my code", "owasp this PR", AI security review, PR review, or changes to AI system code.

The English summary is being prepared. Showing the original description.

Agent skill for security-manager - invoke with $agent-security-manager

The English summary is being prepared. Showing the original description.

Plan and verify security controls for AI agents that read untrusted content and use tools. Review permissions, file boundaries, network destinations, and credentials, then test that rejected actions leave no unintended changes.

• Reviewing security boundaries in an agent runtime
• Checking isolation between tenants and workspaces
• Testing credential cleanup and tool output limits

Evaluate prediction markets as information sources for products, agents, dashboards, and business decisions. Record probabilities implied by market prices with sources and timestamps, assess signal quality, and compare other evidence to recommend suitable uses and explain limitations.

• Evaluating prediction-market data for a product
• Comparing market probabilities with internal dashboard metrics
• Assessing probability signals for agent memory or alerts

Review safeguards for AI agents authorized to sign or send transactions. Cover malicious instructions hidden in external data, spending caps, pre-send simulation, automatic stop conditions, protected transaction routing, and private-key handling through layered security checks.

• Audit a trading bot with transaction authority
• Set per-transaction and daily spending limits
• Design pre-send checks and automatic stops

Design layered protections for AI agents against malicious instructions, tool misuse, and data leakage. Provides guidance and code examples for input checks, isolated tool execution, permission limits, and audit logging to help review and strengthen agent security.

• Review defenses against malicious instructions
• Isolate tool execution and restrict permissions
• Assess data leakage risks between tenants

Build a screening pipeline for web pages, PDFs, and images before an AI agent reads them. Extract hidden text, check for malicious instructions with detection models, and record decisions to block or allow content.

• Screening external content before an agent reads it
• Checking uploaded PDFs and images for hidden instructions
• Scanning documents before adding them to a knowledge base

Audit MCP servers, which provide external tools to AI agents, for hidden instructions, changes after approval, unsafe internal requests, and missing authentication. Use mcp-scan and manual checks to document risks and recommended fixes.

• Reviewing an MCP server before adding it to an agent
• Auditing an internally developed MCP server
• Checking approved tools for unexpected changes

Design security controls for Google Cloud Agent Gateway, covering incoming client requests and outgoing agent tool calls. Produce architecture diagrams, configuration files, deployment commands, and validation scripts using Model Armor, IAP, and Agent Registry.

• Design incoming and outgoing agent communication controls
• Prepare Model Armor filters and tool access policies
• Generate scripts to check authentication and allowed tool calls

Secure AI agents against prompt injection, tool abuse, and data exfiltration with defense-in-depth controls. Use when building, deploying, or hardening agentic AI systems that invoke tools, access data, or interact with production infrastructure.

The English summary is being prepared. Showing the original description.

LLM and AI application security testing — prompt injection, jailbreak resistance, OWASP LLM Top 10 (2025), RAG and agent/tool-use security, model supply chain, and AI red teaming for authorized assessments

The English summary is being prepared. Showing the original description.

Secure AI agents against prompt injection, tool abuse, and data exfiltration with defense-in-depth controls.

The English summary is being prepared. Showing the original description.

Use when the diff calls a model, exposes a tool to one, feeds it fetched or user content, renders or executes its output, stores agent memory or conversation history, asks a user to approve an agent action, routes MCP tools, or changes agent settings, MCP configs, tool policies or prompt files - including this platform's own agents

The English summary is being prepared. Showing the original description.

Test Model Context Protocol (MCP) servers and tool-calling systems for security vulnerabilities including tool injection, parameter manipulation, privilege escalation, and data exfiltration through AI agent tool interfaces. Use this skill when assessing MCP server implementations, AI agent tool integrations, or any system that exposes tools to language models. Covers tool confusion attacks, cross-tool exploitation, and MCP server hardening assessment.

The English summary is being prepared. Showing the original description.

Test AI agent systems for tool abuse, unauthorized actions, privilege escalation through tool chaining, and safety bypass via agentic workflows. Use this skill when assessing autonomous AI agents that use tool-calling (function calling, plugins, actions) to interact with external systems. Covers multi-step attack chains, implicit trust exploitation, and capability boundary testing for AI agents.

The English summary is being prepared. Showing the original description.

Designs, deploys, and secures Google Cloud Agent Gateway solutions. Use when the user needs to configure multi-agent security, ingress (CLIENT_TO_AGENT), or egress (AGENT_TO_ANYWHERE) patterns involving Model Armor, IAP, and Agent Registry. Don't use for general Cloud Load Balancing or basic VPC setup not related to Agent Gateways.

The English summary is being prepared. Showing the original description.

Pre-production security checklist for Agentforce deployments: agent run-as permission scope, grounding-source data classification, action write-scope, audit trail. NOT for the full go-live readiness gate covering cost telemetry, rate limits, rollout and rollback — use agentforce/agentforce-production-readiness-checklist. NOT for a general org security review — use security/security-health-check.

The English summary is being prepared. Showing the original description.

Use for an adversarial red-team / blue-team / auditor review of an AI agent's CONFIG + behaviour (rules, skills, MCP, hooks, permissions) — attack-chain → defensive-gap list, not a code audit.

The English summary is being prepared. Showing the original description.

Use when designing defensive monitoring for agent systems whose risks accumulate across users, sessions, subagents, files, or tool calls. Adds online cross-context clustering with rare escalation, time-to-detect evaluation, provenance checks for control-like workspace content, and sanitized commits for persistent prompt-injection defense.

The English summary is being prepared. Showing the original description.

Detect and defend against prompt injection hidden in documents, web pages, and images consumed by an agent.

The English summary is being prepared. Showing the original description.

Scan Model Context Protocol servers and tool metadata for poisoning, SSRF, and unauthenticated exposure.

The English summary is being prepared. Showing the original description.

Secure AI agents against prompt injection, tool abuse, and data exfiltration with defense-in-depth controls.

The English summary is being prepared. Showing the original description.

Run agent-adapted STRIDE + access-control analysis on an agent system. Produces a ranked risk list with agentic mitigations (scope / split / filter / gate / review).

The English summary is being prepared. Showing the original description.