Act
AISRE
An agentic incident-response pipeline: it detects a live service failure, investigates it with a three-tool orchestrator, and proposes a root cause — but cannot touch anything until a human approves.
What is actually interesting about it
- End-to-end loop: detect a live failure, investigate with a three-tool orchestrator agent, and emit a structured-JSON root cause before any fix is applied.
- Every remediation is gated behind human approval, then automatically health-re-verified after the fix lands.
- A hard zero-financial-API-exposure boundary on all agent tool calls — the agent physically cannot reach a billing surface, rather than being asked not to.
- Seven phases verified end-to-end in a live session: detection, investigation, approval, remediation, re-verification.
- Atlas gates what the agent is allowed to call. KLRB is designed in as the evaluator of its root-cause output. This is the project the other two exist to make trustworthy.
Scope — what this is not
Newest of the three and the least battle-tested — a verified end-to-end pipeline, not something that has carried real on-call load. The approval gate and the capability boundary are the parts I would defend; the detection heuristics are not novel.
Elsewhere in the system
Replaces the shared API key an AI agent uses with a scoped, attenuable token bound to its workload identity — verified offline, with no network hop on the authorization path.
A Kubernetes benchmark that measures whether an LLM actually read the cluster evidence before diagnosing an incident — or just guessed confidently from metadata.
Terraform tells you what drifted. This tells you who changed it, when, and from where — by correlating the plan against CloudTrail.