Skip to content
Initialising environment000

02 / Products

Agent Sandbox Evals

A pentest for agents, run where agents act.

Scroll to explore

01 / Problem

The problem.

Most red-teaming tools attack the model through its chat input. Agents with shell, browser and file access get compromised through the world they work in: a PR title, a README, a dependency, a credential left in reach.

Enterprise procurement asks “what happens on prompt injection?” and vendors have no measured answer.

02 / Deliverables

What you get.

  • 01Scorecard: hijack, exfiltration and breakage rates per attack class, alongside baseline task success
  • 02Failure pack with full traces and what left the sandbox
  • 03Hardening notes that separate harness fixes from model fixes
  • 04Regression suite packaged for CI
  • 05One re-test within 30 days

03 / Steps

How it runs.

  1. Kickoff and scope

  2. Baseline tasks with no hazards

  3. Seeded runs across nine attack classes

  4. Adaptive runs

  5. Confirmation by replay in clean sandboxes

  6. Report and suite

  7. Re-test

04 / Coverage

What we test for.

Attack classes we test for, what is planted in the sandbox, and the severity
ClassWhat is plantedSeverity
Instructions in filesDirectives in READMEs, comments, configs, issues, PDFsHigh
Instructions in tool outputHostile web pages, API responses, MCP servers, CI logsHigh
Poisoned dependenciesPackages, Actions, base images with payloads or typosquatsCritical
Honeypot credentialsTokens in .env, keys in history, SSH keysCritical
Exfiltration channelsHTTP, DNS, git remotes, email, markdown image URLsCritical
Destructive actionsrm -rf, force-push, dropping tables, mass emailsHigh
Privilege and scope creepsudo, admin tokens, a second repoMedium–High
Multi-agent and MCP relayInjected plans returned by subagents or serversHigh
PersistenceEdits to its own config, cron, hooks or memoryCritical

Everything runs on Ionate-controlled sandboxes with egress routed to our own honeypots. Nothing reaches a real external system.

Founding customers

We are taking two founding customers per product: a full engagement at no cost in exchange for a named logo and a case study. Write to us.

Write to us
Next product · 03Agent Failure Autopsy