Skip to content

Incident runbook for the kill-switch #4809

Description

@JSONbored

Context

No written process exists for handling an incident once a misbehaving loop has been stopped.

Note: the kill-switch mechanism itself has been split out into #5670 (Kill-switch implementation), which is now open to Gittensor contributors. This issue covers only the written incident runbook — a maintainer/ops responsibility.

Dependencies

Blocked on #4782 (governance spec) and #5670 (kill-switch implementation, so the runbook can be drilled against a real mechanism).

Requirements

  1. Write an incident runbook covering detection, activation, and post-incident audit-trail review for a misbehaving loop.
  2. Out of scope: the kill-switch mechanism itself — see Kill-switch implementation #5670.

Deliverables / Acceptance Criteria

  • A written, tested runbook
  • A simulated "misbehaving loop" drill is detected and resolved via the runbook within a defined target time

Test Scenarios

Links & Resources


Part of #4778.

Metadata

Metadata

Assignees

Labels

maintainer-onlyOwner-only work — yields no Gittensor points.roadmapOn the Wave-2 agent-layer roadmap board (project 9)

Projects

Status
Todo

Relationships

None yet

Development

No branches or pull requests

Issue actions