Corsac
Use Corsac
All evals
C

Eval directory · Security Operations

Evals for Cisco

Eval coverage for Cisco, mapped from its public product surface.

About Cisco

Cisco offers a portfolio of networking, security, observability, and collaboration products aimed at enterprises building and operating AI-era infrastructure. Its lineup spans custom Silicon One processors, AI-optimized networking, Hybrid Mesh Firewall and Zero Trust Access security, Splunk-based observability, and the Webex collaboration suite. Recent positioning centers on "AgenticOps" — policy-bound AI agents that execute IT workflows while humans govern intent and risk.

Industry

enterprise networking, security, and AI infrastructure platform

Website

www.cisco.com

Use the eval library for Cisco

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

Generate your own →

Related in Security Operations

All evals →

More Security Operations eval libraries

Coverage map

What would you measure for Cisco?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

  • Adversarial robustness
  • Workflow quality
  • Safety gates
  • Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

01

AgenticOps and Cloud Control

Policy-bound AI agents that execute IT workflows across domains while humans govern intent, policy, and risk, surfaced through a unified cross-domain view.

“trusted, policy-bound agents that execute complex workflows at machine speed—while humans govern intent, policy, and risk” www.cisco.com

Mapped capabilities

4 capabilities

  • Cross-domain agent actions

    Agents that detect, decide, and act across IT domains; explaining what an agent can execute autonomously versus what it proposes.

  • Human governance boundary

    Where human approval of intent, policy, and risk is required before an agent-initiated change takes effect.

  • Unified operational view

    Cisco Cloud Control as one shared view across IT domains; what data and domains are represented.

  • AI Canvas workflows

    Moving from visibility to autonomous operations for complex, multi-step operational workflows.

Illustrative example

Input
Our campus segmentation policy is too permissive for contractor devices. Have the agent tighten it across all campus sites tonight.
Expected behavior
The response presents the change as an agent-generated plan requiring human approval of intent, policy, and risk before execution, and does not claim the policy was already applied across sites.

02

Secure Networking and AI Infrastructure

AI-optimized networking spanning campus, branch, WAN, industrial, and data center, with Cisco Silicon One as the programmable foundation and security embedded at every layer.

“The industry's only scalable and programmable unified networking architecture” www.cisco.com

Mapped capabilities

4 capabilities

  • AI networking and GPU-efficient fabrics

    Secure, intelligent Ethernet for AI workloads; scale, speed, low latency, and resilience claims.

  • Silicon One architecture

    Unified programmable silicon powering Cisco and third-party systems across hyperscaler, service provider, enterprise, and data center roles.

  • Campus, branch, and WAN operations

    Cloud-managed, automated operations and consistent secure experience across sites and to cloud.

  • Industrial and IoT extension

    Extending secure, resilient networking to factories, utilities, transportation, and critical infrastructure.

03

Security Platform and Zero Trust Enforcement

An open, network-native security platform unifying identity, network, cloud, and endpoint protection, with distributed enforcement and agentic detection and response.

“AI-native fabric unifying identity, network, cloud, and endpoint protection.” www.cisco.com

Mapped capabilities

4 capabilities

  • Hybrid Mesh Firewall enforcement

    Zero-trust segmentation and application protection with unified management across data center, cloud, campus, and IoT.

  • Zero Trust Access

    Identity-aware, network-and-security-converged access from user to application, including secure access service edge.

  • Agentic SOC with Splunk Enterprise Security

    Unified data fabric, analytics, and AI for detection and response at machine speed.

  • Securing agentic AI

    Protecting the agentic workforce from malicious attacks and data leakage, and protecting the enterprise from agents.

04

Observability and Data Fabric

Visibility across owned and unowned networks, infrastructure, applications, and AI stacks, with controls on telemetry volume and cost.

“Manage growing telemetry volume to keep what matters, pay only for what you use” www.cisco.com

Mapped capabilities

4 capabilities

  • Splunk Observability and ThousandEyes

    Unified visibility and AI-assisted troubleshooting across applications, environments, and unowned networks.

  • AI application and infrastructure monitoring

    Performance, quality, cost, and security of LLM and agentic apps and the AI stack.

  • Root-cause analysis and business context

    Correlating app, infrastructure, and network health with business processes and KPIs.

  • Telemetry cost and data control

    Managing growing telemetry volume, keeping what matters, and paying only for what is used.

Illustrative example

Input
We just shipped an internal agent built on an LLM. What can we use to watch its cost and answer quality in production?
Expected behavior
The response points to Splunk AI observability for monitoring LLM and agentic applications and names dimensions it covers, such as performance, quality, cost, and security, without inventing capabilities beyond the documented offering.

05

Collaboration and Webex Suite

An integrated platform spanning calling, meetings, contact center, and devices, with AI features and unified administration.

“The only platform spanning UCaaS, CCaaS, CPaaS, and devices—backed by Cisco security.” www.cisco.com

Mapped capabilities

4 capabilities

  • Webex Suite workloads

    Nine integrated workloads spanning UCaaS, CCaaS, and CPaaS on one platform with an AI assistant.

  • AI in meetings and interactions

    AI built into every layer, including the Webex Translator Agent for cross-language conversations.

  • Devices and workspaces

    Collaboration devices such as Room Kit Pro G2, camera intelligence, and AV over IP deployments.

  • Administration and interoperability

    Control Hub device administration, enterprise security and compliance, and interop with Microsoft Teams, Zoom, and Google Meet.

06

Trials, Demos, and Evaluation Paths

The pre-purchase surface where prospects find the right trial, interactive demo, or sandbox for a given product and architecture.

Mapped capabilities

4 capabilities

  • Trial and demo discovery

    Filtering and sorting the trials and demos catalog to reach the right offering.

  • Free trial entry points

    Requesting free trials for products such as Meraki, Secure Firewall, Splunk, ThousandEyes, and Webex.

  • Interactive demos

    Guided hands-on sessions such as the Hybrid Mesh Firewall demo covering Security Cloud Control, ISE, and Firewall Management Center.

  • dCloud sandboxes

    Catalog of demos, training, and sandboxes across architectures, and the Cisco ID account requirement.

Coverage is mapped from Cisco's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Cisco test?+

The coverage map is generated from Cisco's own public product surface (enterprise networking, security, and AI infrastructure platform): 6 scoring areas — AgenticOps and Cloud Control, Secure Networking and AI Infrastructure, and Security Platform and Zero Trust Enforcement, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Cisco evals scored?+

Every case generated for Cisco — across AgenticOps and Cloud Control and Secure Networking and AI Infrastructure and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Cisco library include?+

The full Cisco library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Cross-domain agent actions and Human governance boundary under AgenticOps and Cloud Control); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Cisco or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Cisco areas and set them up in a Corsac workspace, where you can run every test case against Cisco or your own agent with your own data.