Skill254 repo starsupdated 2mo ago

sibyl-experiment-supervisor

The sibyl-experiment-supervisor monitors GPU availability on remote machines and automatically dispatches queued machine learning experiments to available hardware. Use this skill to manage large-scale distributed experiment execution, continuously tracking GPU utilization and handling runtime anomalies while distributing compute-intensive tasks across a cluster with configurable polling intervals and allocation thresholds.

View source Repository: AutoResearch-SibylSystem

Install in Claude Code

Copy

git clone --depth 1 https://github.com/Sibyl-Research-Team/AutoResearch-SibylSystem /tmp/sibyl-experiment-supervisor && cp -r /tmp/sibyl-experiment-supervisor/.claude/skills/sibyl-experiment-supervisor ~/.claude/skills/sibyl-experiment-supervisor

Then start a new Claude Code session; the skill loads automatically.

Definition

SKILL.md

!`SIBYL_WORKSPACE="$ARGUMENTS[0]" .venv/bin/python3 -c "from sibyl.orchestrate import render_skill_prompt; import os; ws = os.environ.get('SIBYL_WORKSPACE', ''); print(render_skill_prompt('experiment_supervisor', workspace_path=ws))"`

AGENT_NAME: sibyl-experiment-supervisor
AGENT_TIER: sibyl-standard
SIBYL_ROOT: /Users/cwan0785/sibyl-system

Workspace path: $ARGUMENTS[0]
MODE: $ARGUMENTS[1]
SSH server: $ARGUMENTS[2]
Remote base: $ARGUMENTS[3]
Remote env command: $ARGUMENTS[4]
Task IDs CSV: $ARGUMENTS[5]
Supervisor poll interval sec: $ARGUMENTS[6]
GPU poll interval sec: $ARGUMENTS[7]
GPU free threshold MB: $ARGUMENTS[8]
Max GPUs: $ARGUMENTS[9]
Aggressive mode: $ARGUMENTS[10]
Aggressive threshold pct: $ARGUMENTS[11]