Covers AI engineering, cloud infrastructure, and DevOps — known for google-agents-cli-adk-code, google-agents-cli-observability, and google-agents-cli-deploy.
All Skills by @google
google-agents-cli-adk-code
by @google
ADK code development cheatsheet, providing API references, code patterns, and best practices to accelerate AI Agent development.
google-agents-cli-deploy
by @google
Automate ADK Agent deployment to Cloud Run or GKE, encapsulating Terraform and Docker configurations to simplify production release processes.
google-agents-cli-publish
by @google
Register and publish Gemini Agent to the enterprise marketplace, automatically generating Agent cards, configuring access permissions, and managing versions.
google-agents-cli-workflow
by @google
ADK development workflow guide, standardizing development phases, code reviews, and testing processes to ensure consistent team collaboration.
google-agents-cli-eval
by @google
ADK Agent evaluation framework, automatically running test cases, generating evaluation reports and performance metrics, to verify Agent quality meets standards.
google-agents-cli-scaffold
by @google
Automatically create ADK Agent project structure, generating configuration files, Dockerfiles, and CI/CD templates, supporting new projects and enhancing existing ones.
google-agents-cli-observability
by @google
Configure ADK Agent observability, integrating Cloud Trace, logging, and BigQuery analysis to monitor production environment performance.
bigquery-basics
by @google
This skill provides fundamental knowledge and operational guidance for Google Cloud BigQuery, covering API enablement, dataset and table creation, SQL query execution, command-line tool usage, client library integration, and Infrastructure as Code (Terraform) management. It helps users efficiently perform large-scale data analysis and management on BigQuery, understand its serverless architecture and core concepts, enabling quick adoption of this cloud data platform.
cloud-run-basics
by @google
Run containerized applications on Google Cloud Run, with automatic scaling and no server management.
gemini-api
by @google
Access Google's enterprise-grade AI models via the Gemini API, supporting multimodal understanding and generation.
google-cloud-recipe-auth
by @google
Handle Google Cloud authentication, supporting various authentication methods and credential management.
google-cloud-recipe-onboarding
by @google
This skill provides a streamlined guide for developers to get started with Google Cloud, covering essential steps from account creation and billing setup to project management and deploying their first cloud resource. It simplifies the often complex initial configuration process, helping users quickly launch their cloud projects and accelerate development efficiency.
cloud-sql-basics
by @google
This skill provides fundamental management capabilities for Google Cloud SQL relational database services, covering instance creation, configuration, connection, and security setup. It automates the deployment and maintenance of MySQL, PostgreSQL, and SQL Server databases, significantly boosting development efficiency and ensuring high performance and availability for application data.
google-cloud-waf-security
by @google
This skill generates security-focused guidance for Google Cloud workloads, leveraging the design principles and recommendations from the Google Cloud Well-Architected Framework (WAF) Security pillar. It helps evaluate workloads, identify security requirements, and provides actionable recommendations across IAM, network security, data protection, and operational security, enabling users to build a robust cloud security posture.
google-cloud-waf-cost-optimization
by @google
This skill generates cost optimization guidance for Google Cloud workloads, leveraging the Google Cloud Well-Architected Framework. It evaluates workloads, identifies cost requirements and constraints, and provides actionable recommendations to efficiently build, deploy, and manage workloads on Google Cloud. This leads to continuous cost savings and maximized business value.
google-cloud-waf-reliability
by @google
This skill generates reliability-focused guidance for Google Cloud workloads, drawing from the Google Cloud Well-Architected Framework's reliability pillar. It assists users in evaluating workloads, identifying specific reliability requirements, and providing actionable recommendations to build, deploy, and manage systems reliably on Google Cloud. This ultimately minimizes downtime, enhances user experience, and ensures data integrity.
gke-basics
by @google
This skill helps users plan, create, and configure production-ready Google Kubernetes Engine (GKE) clusters, with a strong emphasis on the golden path Autopilot configuration. It covers critical aspects from Day-0 checklists, Autopilot vs. Standard comparisons, to networking, security, observability, scaling, cost optimization, and AI/ML inference. The goal is to simplify the deployment and management of GKE environments, ensuring efficient and secure cluster operations.
alloydb-basics
by @google
The AlloyDB Basics skill provides fundamental management capabilities for AlloyDB for PostgreSQL databases on Google Cloud, including cluster and instance creation and configuration. It helps developers leverage AlloyDB's high-performance, high-availability features and integrate its AI functionalities (such as vector search and AI functions) to accelerate the development of enterprise-grade AI applications, optimizing data storage and processing efficiency.
google-cloud-networking-observability
by @google
This skill serves as a Google Cloud Networking Observability Expert, adept at quickly analyzing VPC Flow Logs, Firewall Logs, Threat Logs, and network metrics to identify and resolve network issues. It streamlines troubleshooting by efficiently querying data and applying predefined guides, enabling rapid fault localization, performance trend monitoring, and configuration verification. This significantly boosts network operations efficiency and accelerates problem resolution.
gcloud
by @google
This skill enables AI agents to safely and efficiently manage Google Cloud resources using the gcloud CLI. It enforces syntax validation, parameter verification, and dry-run checks to prevent command hallucinations and destructive actions. Ideal for provisioning, configuring, and querying cloud resources with improved accuracy and speed.
agent-platform-skill-registry
by @google
This skill provides comprehensive tools for interacting with the Skill Registry on the Gemini Enterprise Agent Platform. It enables users to efficiently discover, manage the lifecycle of, and monitor operations related to agent skills. Developers can easily search, upload, update, and delete skills, ensuring a well-organized and maintainable skill ecosystem for AI agents. Additionally, it supports scaffolding new agent skills, streamlining the development process.
google-cloud-waf-operational-excellence
by @google
This skill provides recommendations for efficiently operating workloads on Google Cloud, based on the Operational Excellence pillar of the Google Cloud Well-Architected Framework. It covers core principles such as operational readiness, incident management, resource optimization, automation, and continuous improvement. This helps users design, implement, and manage high-performing, secure, and reliable cloud solutions, ensuring adaptability to evolving cloud needs.
gemini-agents-api
by @google
This skill provides the Managed Agents API for the Gemini Enterprise Agent Platform, enabling developers to programmatically create, configure, list, update, and delete stateful, server-managed agent resources. It acts as the control plane for the platform, allowing for tailored agent containers equipped with system instructions, sandboxed files, custom skill registries, and local/remote tools. This significantly streamlines the lifecycle management of AI agents, accelerating their development, deployment, and operational efficiency.
google-cloud-waf-performance-optimization
by @google
This skill, based on the Performance Optimization pillar of the Google Cloud Well-Architected Framework, offers principles and recommendations for designing, building, and operating high-performing cloud workloads. It focuses on efficient resource allocation, leveraging elasticity, promoting modular design, and continuous monitoring for improvement. This helps users achieve excellent performance and cost-effectiveness on Google Cloud, ensuring stable and efficient system operation.
google-cloud-waf-sustainability
by @google
This skill leverages the Sustainability pillar of the Google Cloud Well-Architected Framework, offering principles and recommendations to help users optimize their cloud workloads for reduced environmental impact. It guides users in selecting low-carbon regions, optimizing resource usage, developing energy-efficient software, and continuously measuring and improving, thereby significantly lowering their carbon footprint and enhancing energy efficiency.
agent-platform-deploy
by @google
This skill offers comprehensive guidance for deploying and managing open models from the Agent Platform Model Garden to endpoints. It covers model copying, deployment, resource cleanup, and critical safety protocols for operations, ensuring the secure and efficient online deployment and maintenance of AI models.
agent-platform-rag-engine-management
by @google
This skill offers comprehensive guidance on interacting with the Agent Platform RAG Engine. It enables developers to efficiently manage RAG engines using the Python SDK, covering operations such as listing and inspecting corpora, retrieving contexts, and generating answers based on retrieved information. This capability is crucial for building and maintaining powerful knowledge storage and retrieval systems for AI agents, thereby improving their decision-making and response accuracy.
agent-platform-tuning
by @google
This skill provides procedural knowledge for fine-tuning Large Language Models (both Open Models and Gemini Models) using Agent Platform's tuning service. It covers the entire lifecycle from environment setup and data preparation to job configuration, monitoring, and deployment, enabling users to optimize model performance for specific application needs.
agent-platform-prompt-management
by @google
This skill provides the Google Agent Platform SDK for efficient AI prompt management. It supports creating, listing, retrieving, and deleting prompts, along with version control and best practices. This helps developers standardize and optimize large language model prompts, enhancing the performance and consistency of AI applications.
agent-platform-tuning-management
by @google
This skill provides instructions on managing GenAI Tuning Jobs using the Agent Platform Python SDK. It enables users to check the status of tuning runs, find active jobs, or cancel long-running jobs, thereby streamlining AI model development, optimizing resource utilization, and ensuring efficient project execution.
agent-platform-eval-flywheel
by @google
This skill helps users evaluate and iteratively improve GenAI models and agents using the Agent Platform GenAI Evaluation SDK. It supports creating evaluation datasets from session traces, DataFrames, or synthetic generation, selecting and configuring custom metrics, analyzing rubric verdicts and loss patterns, and suggesting concrete code/prompt improvements based on eval results, enabling a quality flywheel of continuous improvement.
agent-platform-model-registry
by @google
This skill provides commands to manage machine learning models in the Agent Platform Model Registry, including listing, describing, uploading, updating, and deleting models. It helps developers efficiently manage model lifecycles, ensure version control, and prevent accidental destructive actions through safety confirmation tiers.
agent-platform-endpoint-management
by @google
This skill provides procedural knowledge for managing Agent Platform Endpoints, enabling users to efficiently create, update, delete, and query online prediction endpoints for AI models. It simplifies the deployment and management of cloud-based AI infrastructure, ensuring stable operation of AI services and strict adherence to operational safety protocols.
agent-platform-inference
by @google
This skill provides comprehensive instructions for developers to authenticate and connect to Google Cloud Agent Platform, enabling the use of both First-Party (Gemini) and Third-Party (OpenMaaS) Generative AI models for inference. It guides users through critical environment setup steps, including Google Cloud authentication, API enablement, virtual environment creation, and dependency installation. By streamlining the connection process, this skill ensures developers can efficiently integrate and leverage advanced AI capabilities in their applications, while also emphasizing safety tiers for inference execution to prevent unexpected costs.
agent-platform-migrate-from-ai-studio
by @google
This skill facilitates the seamless migration of applications from the developer-centric Google AI Studio to the enterprise-grade Google Cloud Agent Platform. By leveraging this skill, users can apply GCP credits towards Gemini API costs, unify inferencing pipelines, IAM permissions, telemetry, and billing with existing Google Cloud infrastructure. It also provides enterprise-level data privacy, security compliance, and MLOps ecosystem support, enhancing the stability, security, and scalability of AI applications.
google-analytics-data-api-basics
by @google
This skill provides programmatic access to Google Analytics report data via the Data API, enabling automated reporting, custom dashboards, and enterprise integration. It empowers analysts and developers to streamline data extraction and focus on insights.
google-analytics-admin-api-basics
by @google
This skill provides a getting started guide for the Google Analytics Admin API, enabling developers to automate account and property management. It allows users to configure data streams, custom dimensions, and integrations programmatically, saving time and reducing manual errors.
workload-manager-basics
by @google
This Skill leverages Google Cloud Workload Manager to automatically validate enterprise workloads against GCP best practices and recommendations. Users define a resource scope, select built-in or custom rules, run an evaluation, and inspect findings to identify compliance issues, reduce risk, and optionally export results to BigQuery for deeper analysis.
bigquery-ai-ml
by @google
This skill enables direct invocation of Vertex AI models within BigQuery using built-in functions like AI.FORECAST, AI.GENERATE, AI.DETECT_ANOMALIES, and AI.KEY_DRIVERS, allowing time-series forecasting, anomaly detection, key driver analysis, and content generation directly in SQL. It eliminates the need to move data or build custom ML pipelines, accelerating data analysis and AI adoption. Ideal for analysts and data scientists seeking to enrich their queries with ML without leaving the BigQuery environment.
agent-platform-alert-configuration
by @google
This skill provides dynamic threshold alerting configurations for AI agents on Google Cloud Vertex AI, using 1-week lookback baselines via PromQL. It reduces alert noise from static thresholds and covers reliability (latency, error rate) and quality (response quality, hallucination) monitoring, ensuring stable agent operations.
bigtable-basics
by @google
This skill helps users efficiently manage Google Bigtable, covering instance configuration, data querying, schema design, and performance optimization. It avoids common pitfalls like full table scans, significantly improving operational and development efficiency.
gke-inference
by @google
This Skill enables rapid deployment of AI/ML inference workloads on GKE by auto-generating optimized Kubernetes manifests and selecting the best GPU/TPU accelerators, greatly simplifying the deployment and scaling configuration for LLM models such as Llama, Gemma, and Mistral.
gke-upgrades
by @google
This skill produces customized upgrade plans, runbooks, and checklists tailored to your GKE cluster environment. It automatically gathers cluster mode, version, release channel, and workload types to deliver safe, efficient upgrade strategies, helping teams avoid disruptions and maintain production stability.
gke-observability
by @google
This Skill enables users to configure comprehensive GKE observability with one command, including monitoring, logging, Managed Prometheus, and Dataplane V2 metrics, helping quickly diagnose control-plane and workload issues.
gke-networking
by @google
This Skill provides best-practice guidelines for GKE cluster networking, covering private clusters, VPC-native mode, IP planning, and more, enabling users to quickly set up secure and efficient Kubernetes networking with fewer misconfigurations and less troubleshooting.
gke-golden-path
by @google
This skill automatically applies the recommended Golden Path configuration for GKE production clusters, simplifying cluster creation and setup while ensuring best practices, with support for customizations and trade-off documentation.
gke-cluster-creation
by @google
This skill provides best-practice templates and mode selection (Autopilot vs Standard) for creating Google Kubernetes Engine clusters, covering security, cost optimization, and high availability. Users can quickly generate compliant cluster configurations without consulting scattered documentation, reducing manual error risks.
gke-storage
by @google
This skill helps GKE users quickly configure and manage Kubernetes storage resources, including PersistentVolumeClaims, StorageClasses, volume expansion, and best practices, improving storage deployment efficiency and reducing configuration errors.
gke-reliability
by @google
This skill configures and validates GKE cluster high availability, including regional clusters, Pod Disruption Budgets, health probes, and topology constraints, helping teams build resilient workloads and reduce failure impact.
gke-app-onboarding
by @google
This Skill provides a complete workflow for containerizing and deploying applications to GKE for the first time, covering app assessment, Docker image building, image management, Kubernetes manifest generation, and deployment verification. It helps development teams migrate apps to GKE quickly and consistently, reducing manual errors and improving deployment efficiency.
gke-compute-classes
by @google
This skill provides guidance on configuring, optimizing, and troubleshooting GKE ComputeClasses for cost optimization (e.g., Spot VMs with on-demand fallback), GPU/TPU workloads (targeting specific accelerators), performance tuning (selecting specific machine families), and zone targeting (colocating workloads with zonal resources). It helps users quickly generate best-practice YAML templates and avoid common pitfalls.
gke-backup-dr
by @google
Protect stateful GKE workloads with automated backup and disaster recovery using Backup for GKE. This skill captures Kubernetes resource metadata (manifests, configurations, secrets) and underlying persistent volume data. It enables creating backup plans with retention and cron schedules, triggering manual backups, and executing restores to target clusters with configurable conflict policies. Ideal for ensuring data integrity, meeting compliance requirements, and enabling rapid recovery in production GKE clusters. The skill also provides best practices for CMEK encryption, namespace scoping, application consistency, and CSI volume snapshots.
google-agents-cli-onboarding
by @google
This skill initializes the Google Agents CLI toolchain and guides developers through the full lifecycle of building, evaluating, deploying, and publishing AI agents. By automating commands and providing structured workflows, it reduces the learning curve and boosts multi-phase collaboration efficiency.
gke-multitenancy
by @google
This skill automates the configuration and management of multi-tenancy patterns on GKE, including namespace isolation, RBAC planning, resource quotas, and network segmentation. It helps teams achieve secure isolation, cost allocation, and least-privilege access on shared clusters, reducing operational complexity.
gke-batch-hpc
by @google
This skill helps users run batch processing and high-performance computing workloads on Google Kubernetes Engine efficiently, providing complete guidance from Job configuration and MPI workloads to cost optimization (Spot VMs, scale-to-zero), significantly simplifying deployment and operations.
bigquery-bigframes
by @google
This skill leverages BigFrames (BigQuery DataFrame) API to process large-scale data efficiently in the cloud, supporting data cleaning, transformation, analysis, and machine learning without moving data locally. It brings the power of BigQuery to Python developers with familiar Pandas syntax, eliminating OOM errors and accelerating workflows.
detection-engineering-coverage-evaluation
by @google
This skill automates the end-to-end detection engineering lifecycle, including threat intelligence extraction, generating Threat Detection Opportunities (TDOs), creating synthetic events, evaluating rule coverage, auditing rule status, and generating new rules to fill gaps, helping security teams systematically improve detection coverage and response efficiency.
google-cloud-recipe-foundation-builder
by @google
This skill automates the deployment of a secure, enterprise-grade Google Cloud landing zone, setting up security guardrails, resource hierarchy, and centralized logging/monitoring, helping teams quickly establish a compliant cloud foundation.
datalineage-bigquery-asset-impact-analysis
by @google
This skill performs downstream impact analysis (blast radius assessment) for BigQuery tables or views, automatically discovering dependent ETL pipelines, views, and queries. It helps users evaluate the scope of risk before making changes or performing maintenance, preventing unintended disruptions.
adk-dev-guide
by @google
ADK Development Guide: Read at the start of an ADK agent development session. AI Agent Skill, improves work efficiency and automation capabilities.
adk-cheatsheet
by @google
Provides a cheatsheet that must be read before writing or modifying ADK agent code. Covers key information and best practices for the ADK API, ensuring developers complete tasks efficiently and accurately.
adk-deploy-guide
by @google
A deployment guide that must be read before deploying any ADK agent. Details the ADK agent's deployment process, configuration requirements, and best practices to ensure smooth and efficient deployment.
adk-scaffold
by @google
A scaffolding guide that must be read before creating or enhancing any ADK agent project. Provides best practices for project structure, initialization, and development workflow, ensuring projects are standardized and efficient.
adk-observability-guide
by @google
A guide that must be read before setting up observability for ADK agents. Covers configuration and practices for logs, metrics, and tracing, ensuring transparent system status for monitoring and troubleshooting.
adk-eval-guide
by @google
An evaluation guide that must be read before running any ADK assessment. Details the ADK assessment process, standards, and tools to ensure accurate and reliable results, providing a basis for decision-making.
mantis-threat-model
by @google
This skill synthesizes trust boundaries, attack surfaces, and attacker profiles exclusively from the architecture and entity definitions stored in a knowledge base, then writes a structured THREAT_MODEL.md document carrying a snapshot provenance header. It turns ad-hoc security review into a reproducible, idempotent workflow: the prior model is archived per pass, stale snapshots are flagged, and every conclusion stays traceable to a specific KB snapshot. Use it during architecture reviews, design changes, or recurring security re-checks where consistency and auditability matter.
mantis-reflect
by @google
During multi-agent runs, this Skill parses each stage's execution trajectory — thoughts, tool calls and observations — and distils reusable insights into structured entries appended to a shared learnings file, de-duplicating on semantic content and stamping each entry with snapshot provenance. It turns one agent's mistake into durable team memory so later agents stop repeating the same failed calls and wasted steps. Missing or unreadable transcript logs are recorded as explicit insights instead of silently yielding zero learnings, keeping the improvement loop trustworthy across harnesses.
mantis-pipeline-adapter
by @google
This Skill is an interactive pipeline-design consultant that helps developers build their own deterministic orchestrator harness for multi-stage agent skills. It asks about the target language, framework, execution environment and scale, then recommends architecture patterns such as a single-source-of-truth state store, UUID-based referencing for token efficiency, strict inter-stage schema contracts, locking for concurrent runs, and optional snapshot, retrieval and SAST seeding extensions. It is useful when a team needs a reliable, scalable and auditable agent pipeline rather than ad-hoc prompts.
mantis-meta-agent
by @google
This Skill acts as a persistent supervisor that launches and monitors long-running automated code review campaigns, tracking pass state, the findings ledger, and immutable snapshots, and verifying each finding's state before handing off between subagents. It lets multi-pass, multi-agent review workflows resume after interruption and stay fully auditable instead of restarting from scratch.
mantis-review
by @google
mantis-review is an independent validator skill that adversarially re-checks deduplicated findings against the real source code, assuming every claim is a false positive until proven otherwise. It filters out noise and hallucinations, then writes the verdict (status, reasoning, reproduction hints) back to disk. It fits the triage stage after automated scanning, turning long suspect alert lists into a short set of trustworthy issues.
mantis-summarize
by @google
This skill walks a repository and writes a security-focused, deterministic mantis-summary.md into every source directory, rolling child summaries up into their parents. Downstream planning and research stages read these summaries instead of raw source, so context size and token cost drop sharply while the analysis stays reproducible. It is useful when an agent must audit or navigate a large codebase under tight context limits and needs a stable, idempotent map of what each directory contains.
mantis-architecture
by @google
This skill synthesizes ephemeral insights from a learnings queue and structural analysis of a codebase into a canonical, interlinked Markdown knowledge base: it defines the system architecture, maps concrete entities and components, and categorizes historical vulnerability patterns under workspace/kb. Every rewritten file is stamped with a snapshot marker and archived per pass, so later agents reuse one durable, traceable architecture memory instead of re-deriving context, and the knowledge base stays provably fresh against a pinned snapshot.
mantis-structural-index
by @google
This Skill builds a content-addressed semantic-unit index for a codebase and exposes caller lookup, function-boundary retrieval, and call-site awareness, persisting results as a manifest plus a query-optimized catalog. It reuses units across snapshots for incremental rebuilds, attaches prebuilt SCIP/Kythe/LSIF indexes, and degrades gracefully to text search while staying read-only on source. It never hard-fails—if tools or parsing are unavailable it returns an empty index—so developers and AI agents can map call chains and dependencies quickly before reading code.
mantis-critic
by @google
This Skill re-verifies each validated security finding against the pinned source snapshot and decides whether the flaw is still triggerable in a compiled, optimized release build, dropping debug-only features, assertion traps, and unreachable paths. It fits right before severity rating and backlog assignment, so teams spend remediation effort only on issues that actually reproduce in production.
mantis-plan
by @google
This Skill acts as a security architect that maps a codebase's external boundary and produces an adaptive, threat-model-driven defensive review plan. It analyzes directory metadata, historical findings, and version-control diffs to decide whether a full scan or an incremental review is needed, then re-schedules unresolved issues for fresh discovery. Use it when you need a repeatable, evidence-based auditing roadmap instead of an ad-hoc checklist.
mantis-researcher
by @google
This skill performs rapid triage and deep-dive reviews of production source files to surface missing boundary checks, weak preconditions, unsanitized input, and interface violations. It turns a manual file-by-file audit into structured, deduplicated, and traceable findings driven by an audit plan, so security coverage scales across large codebases.
mantis-report
by @google
This skill synthesizes complex, technical vulnerability finding logs into a high-quality, human-readable security review packet for developers and stakeholders. It consolidates confirmed findings and exploit chains across current and archived passes, includes snapshot provenance, and writes pass-tagged Markdown reports with an always-current latest copy for remediation handoff.
mantis-dedupe
by @google
This Skill clusters and consolidates raw security scan findings: identical or heavily overlapping issues are merged into a single descriptive record with unified fields and history, duplicates are archived with status DUPLICATE, and uncertain matches get a soft possible_duplicate_of tag. It also filters out findings already triaged in earlier passes and flags cross-snapshot matches as possible regressions. Use it after every scanning round so reports stay short, trustworthy and free of repeated noise.
mantis-reproduce
by @google
This Skill acts as an integration test engineer for security findings: it designs crash reproducers or attack inputs for each discovered flaw and executes them inside an isolated sandbox to empirically confirm the bug. It records the reproducer file, run command, and raw output against the exact code snapshot, and supports re-attack verification after a patch is applied, so verdicts stay reproducible and auditable.
mantis-calibrate
by @google
This Skill acts as a risk calibration expert: it reads confirmed findings, the threat model and pipeline state, then scores each issue on a rigorous 1-10 risk matrix that factors in reproduction success and production viability. It writes impact, likelihood, availability tier, priority, calibration checklist and an executive summary back into the finding files, giving security and engineering teams a consistent, reproducible way to rank what to fix first.
mantis-history
by @google
This Skill analyzes a repository's version control history to automatically extract past vulnerabilities, security fixes, patches, and related files, producing a structured historical learnings database for downstream use. It removes the pain of manually digging through commit logs and gives security audits and vulnerability retrospectives reusable historical context.
mantis-patch
by @google
This Skill generates minimal, correct code fixes for already-reproduced security flaws, applies them transactionally inside shadow directories or file backups, and re-attack verifies each patch inside isolated sandboxes, rolling back automatically on failure. It makes remediation reproducible and auditable, reduces the risk of breaking existing business logic, and records each patch outcome in long-term memory for reuse on similar vulnerabilities.
mantis-chain
by @google
This Skill is a vulnerability chainer that correlates isolated, individually-validated security findings with knowledge-base primitives to construct multi-step exploit chains. It deduplicates chains by finding signature plus code-path tiebreak across passes, so security teams can surface genuinely exploitable high-risk attack paths hidden inside scattered low-severity issues.