Incident Response
Your agent did something costly or public. We reproduce it, test the blast radius and verify the fix.
The problem
You need to know quickly how widespread the failure is and whether the fix holds.
What we do
- Reproduction attempts and variant testing
- Fix verification when your fix ships
- Permanent regression tests
What you receive
- A first findings note within 48 hours of access
- A final report within 5 business days
What you provide
- Incident description
- Transcripts under a data agreement
- Written authorization
Limits
- Starts only after our formal data policy is published and a data agreement is signed, because it uses your transcripts or outputs (outside-in reproduction can start sooner)
- Outside-in testing is limited to 60 messages per day (about 120 in 48 hours)
- One incident at a time; accepted when capacity allows
- No communications or legal advice
Results describe behavior within the tested scope and configuration. They give you independent evidence for security reviews, audits and compliance work; certification itself is issued by accredited bodies.
Availability and independence
Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms. Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.
Pilot deliveries are scoped in writing, tell you it is an early delivery, and follow the same review and evidence rules as any engagement.
Independent evaluation: we had no hand in building or advising on your agent, and our price is fixed before testing starts, so the results don't change what we're paid.
What is covered
The test areas this service always includes, with their status at the access level it uses.
- Incident reproduction and variant testingHow bad is it, and is it fixed?Pilot: Procedure designed and deliverable with existing tools, but not yet delivered to a client. Sold under pilot terms.
- Post-incident regression casesWill it happen again?Available now: Built into the TrendLab evaluation engine and used in sealed runs.