Autonomous Ops Agent
Monitor fake metrics, diagnose incidents, propose actions, require HUMAN_APPROVAL before anything destructive, and write an incident report.
Outcome: An ops agent that detects SLO burn, proposes a bounded action plan, blocks destructive tools without HUMAN_APPROVAL, and emits a post-incident report.
advanced · 8–12 hours
- 01
Overview and Architecture
Define an ops agent as observe-metrics → diagnose → propose → gated act → report, with never-destructive-by-default as a hard invariant.
- 02
Metrics World and Read Tools
Simulate services, SLOs, logs, and read-only tools that never mutate world state.
- 03
Diagnose and Propose Actions
Map symptoms to a diagnosis object and a list of proposed actions with risk, destructiveness, and expected effect — without executing them.
- 04
HUMAN_APPROVAL Gate
Implement execute() so every mutate tool checks an approval map, records attempts, and never changes the world on deny or missing approval.
- 05
Incident Report and Hardening
Emit a timeline report, re-check SLOs after gated actions, and eval that destructive tools never fire without approval.