google

Architecture notes often leave attacker access and trust boundaries unclear. This skill turns existing Mantis knowledge into a structured threat model with deployment intent and snapshot provenance.

Large source trees make symbol navigation and caller discovery expensive. This skill guides structural indexing with reusable caches and coverage-aware queries while preserving lexical discovery.

Security fixes can fail when verification uses stale code or broken test harnesses. This skill guides minimal patches with snapshot checks, baseline tests, and independent re-attacks.

Long security reviews require coordinated specialist stages and reliable campaign state across repeated passes. This skill supervises Mantis subagents, preserves findings, and manages optional synchronization and snapshot provenance.

Past security fixes are difficult to track across repository history. This skill guides an AI agent to extract, classify, and cache historical vulnerabilities for later analysis.

Repeated security findings increase review effort and can obscure regressions. This skill guides consolidation while preserving provenance, active regression candidates, and transaction history.

Security findings need consistent priorities grounded in exploit evidence and architectural context. This skill applies a risk matrix and documented caps to existing Mantis findings.

ADK 智能体项目需要在脚手架、测试、部署和运维之间完成许多相互关联的步骤。此技能提供分阶段的 agents-cli 工作流，帮助编码智能体一致地遵循这些步骤。

启动一个 Google ADK 智能体项目需要在模板、部署和工作流方面做出多项选择。此技能会指导基于 agents-cli 的脚手架搭建，从原型设置到增强和升级规划。

将 ADK 智能体发布到 Gemini Enterprise 需要正确的注册模式、资源 ID 和权限。此技能会指导 Claude、Codex 或 Claude Code 安全地完成 CLI 注册和 Agent Registry 管理。

生产环境 ADK 智能体需要跟踪、日志和分析信号，同时不暴露提示词数据。此技能指导 Cloud Trace、提示词日志记录、BigQuery 分析以及提供商选择。

代理团队在更改提示词、工具或部署设置之前，需要可重复的证据。此技能指导 Google ADK 代理的数据集设计、指标选择、评分、比较和失败分析。

在 Google Cloud 目标上部署 ADK 代理需要做出许多基础设施选择。此技能会指导目标选择、部署命令、密钥、CI/CD、测试和恢复。

开发者在构建 Python agents 时需要可靠的 ADK 模式。本技能为 tools、callbacks、state、多 Agent 编排和 workflow 代码提供简明参考。