Skip to content
Get packs

Oversight Team Monitor

Lets an agent act as a monitor seat that watches other AI rigs for drift and intervenes lightly.

Source Download skill
Repository
mvschwarz/openrig
Stars
8,553
Licence
Apache-2.0
Files
1 file

What it does

This skill teaches an AI agent how to act as a monitor on an oversight pod, a standing team of agents whose job is to keep other agent teams (rigs) healthy. The agent stays idle until a warning flag fires, then checks evidence such as transcripts, git history, and queue sizes for patterns like premature parking, process drift, off-task wandering, or runaway token use. It then corrects the problem with the lightest touch: first pinging the other rig's orchestrator, then a refocus reminder, and only escalating to a human when needed.

Great for

  • Developers running multi-agent rigs in the OpenRig framework who want a supervisor seat that watches other rigs
  • Teams designing cost-conscious monitoring where cheap models summarize logs and an expensive model makes decisions

Skip it if

  • Anyone wanting to run or orchestrate their own rig; the skill explicitly says that is the orchestration-team skill instead
  • Readers without an OpenRig setup: the detectors and intervention steps depend on that framework's commands and sibling skills, and the file is marked as a draft

Try asking

  • “Use the oversight-team skill to check whether the rig in pod alpha has been idle with an unfinished handoff, and confirm from its transcript before intervening.”
  • “Configure the v0 drift detectors across my fleet and tell me which seats show token burn.”
  • “Should I ping this pod's orchestrator or escalate to the human? Walk me through the intervention ladder for this drift case.”
Step-by-step guide: how to use Oversight Team MonitorInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.

    oversight-team.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip oversight-team.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip oversight-team.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Read the referenced sibling skills (watchdog, refocus, retiring-and-inheriting-a-seat) before relying on this one, since it defers mechanics to them.
  • Verify available telemetry against your OpenRig version: the file warns that per-agent token tracking over time is not assumed to exist and the v0 detector uses point-in-time polling.
  • Respect the swap-window suppression rule so detectors do not flag seats that are mid-handover.

What's inside

  • SKILL.md

Redistributed unmodified from mvschwarz/openrig under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.