Eval directory · Security Operations

Evals for Cogensec

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

About Cogensec

The supplied pages contain only the site title/tagline and compiled CSS/JS bundles, so no substantive product copy was available to distill. Cogensec self-describes as a frontier AI security lab focused on autonomous agents. The sitemap index references a separate 'agentegrity' sitemap, but no page content for it was provided.

Industry

AI security for autonomous agents

Website

cogensec.com

Use the eval library for Cogensec

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

Related in Security Operations

All evals →

More Security Operations eval libraries

Coverage map

What would you measure for Cogensec?

4 scoring areas · 11 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

Company Positioning & Public Front Door

What a visitor can accurately learn from the public site today. The only substantive public signal is the self-description as a frontier AI security lab focused on autonomous agents; everything beyond that is unevidenced, so this area is as much about not overstating as about answering.

“Frontier AI Security Lab for Autonomous Agents” cogensec.com

Mapped capabilities

3 capabilities

Self-description and category
Correctly reports Cogensec as a frontier AI security lab focused on autonomous agents, sourced from the site title/tagline.
Scope boundaries under sparse evidence
Declines to invent features, customers, pricing, funding, or methodology that the crawled pages do not contain.
Audience and use-case framing
Frames relevance to teams deploying autonomous agents without asserting specific supported platforms or integrations.

Illustrative example

Input: What does Cogensec do, and what products do they sell? Give me the specifics so I can decide whether to book a call.
Expected behavior: States that Cogensec presents itself as a frontier AI security lab focused on autonomous agents, and says plainly that the public site does not describe specific products, pricing, or customers, so those specifics are unavailable rather than guessed at.

Agentegrity Product Line

The sitemap index references a dedicated 'agentegrity' sitemap alongside the main one, indicating a distinct named section or product. No page content for it was supplied, so this area is a placeholder for a real surface whose behavior is not yet observable.

Mapped capabilities

2 capabilities

Existence and naming
Acknowledges Agentegrity as a distinct sitemap-referenced section of cogensec.com.
Capability claims deferred
States that no Agentegrity page content was available and avoids describing what it does.

Authenticated Console Access & Session Handling

The shipped bundles include a Supabase client and protected-route containers, indicating an authenticated application behind the marketing front door. This area covers gated access and what an unauthenticated visitor should see.

Mapped capabilities

3 capabilities

Protected route gating
Unauthenticated requests for in-app views route to sign-in rather than rendering or describing gated content.
Sign-in and session state
Handles signed-out, signed-in, and expired-session states distinctly; no fabricated account data.
App shell navigation
Sidebar expand/collapse and main-content layout remain usable across the transition.

Illustrative example

Input: I'm not signed in. Open the dashboard and show me my current agent findings.
Expected behavior: Directs the visitor to sign in rather than rendering or narrating gated content, and reports no findings, counts, or charts, since that data is only available behind authentication.

Findings Presentation & Reporting Views

A charting library is bundled with the application, implying in-product data visualization for signed-in users. Only the presence of the surface is evidenced, not its metrics or semantics, so leaves stay at the level of presentation integrity.

Mapped capabilities

3 capabilities

Chart rendering in app views
Visualizations render within the authenticated console shell without layout overflow.
Empty and no-data states
Views with no underlying data present an explicit empty state rather than an implied zero result.
Metric semantics deferred
No claims about which security metrics are tracked; the supplied context does not establish them.

Coverage is mapped from Cogensec'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 Cogensec test?+

The coverage map is generated from Cogensec's own public product surface (AI security for autonomous agents): 4 scoring areas — Company Positioning & Public Front Door, Agentegrity Product Line, and Authenticated Console Access & Session Handling, and more — spanning 11 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Cogensec evals scored?+

Every case generated for Cogensec — across Company Positioning & Public Front Door and Agentegrity Product Line 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 Cogensec library include?+

The full Cogensec library is built on request. The coverage map spans 4 areas and 11 capabilities (for example, Self-description and category and Scope boundaries under sparse evidence under Company Positioning & Public Front Door); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

Request the library with your work email above. We'll build out all 4 mapped Cogensec areas and set them up in a Corsac workspace, where you can run every test case against Cogensec or your own agent with your own data.