「tool-abuse」の検索結果

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.

日本語の概要は準備中です。原文の説明を表示しています。

Detect tool misuse and unexpected code execution via dialogue testing. Use when the agent exposes file, code-execution, or network tools.

日本語の概要は準備中です。原文の説明を表示しています。

Secure the path from an inbound event (webhook, email, SMS, GitHub comment, file, cron) to an agent spawn in the Port Daddy fleet. Use when wiring trigger sources to agents, reviewing the io-wiring registry, adding a trust/permission gate, hardening output sinks (SSRF/path traversal), scoping tools by provenance, or proving the spawn path is safe against prompt-injection → tool-abuse. Encodes ADR-0093 and the red-team threat model. NOT for generic Coast Guard sandboxing (ADR-0050), the macaroon push-grant gate alone (ADR-0053), or non-fleet code.

日本語の概要は準備中です。原文の説明を表示しています。