Menü
Validate HPC environment before experiment deployment. Checks GPU, rendering, network, dependencies, and model loading. Use before submitting SLURM jobs.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
| name | preflight-hpc |
| description | Validate HPC environment before experiment deployment. Checks GPU, rendering, network, dependencies, and model loading. Use before submitting SLURM jobs. |
| argument-hint | ["server-name"] |
| allowed-tools | Bash(*), Read, Write |
Validate compute environment on: $ARGUMENTS
ssh $SERVER "echo 'SSH OK'; hostname; nvidia-smi -L | head -2"
ssh $SERVER "nvidia-smi --query-gpu=index,name,memory.used,memory.total --format=csv,noheader"
ssh $SERVER "env | grep -i proxy; curl -s --max-time 5 https://huggingface.co > /dev/null 2>&1 && echo 'HF reachable' || echo 'HF BLOCKED — set HF_HUB_OFFLINE=1'"
ssh $SERVER "conda activate <your-conda-env> 2>/dev/null && python -c 'import torch; print(f\"PyTorch {torch.__version__}, CUDA: {torch.cuda.is_available()}\")'"
ssh $SERVER "srun --gres=gpu:1 --time=5:00 --partition=<your-partition> bash -c '
export MUJOCO_GL=egl
python -c \"
import mujoco
m = mujoco.MjModel.from_xml_string(chr(60)+\"mujoco\"+chr(62)+chr(60)+\"worldbody\"+chr(62)+chr(60)+\"light/\"+chr(62)+chr(60)+\"geom type=\\\"plane\\\" size=\\\"1 1 .1\\\"/\"+chr(62)+chr(60)+\"/worldbody\"+chr(62)+chr(60)+\"/mujoco\"+chr(62))
d = mujoco.MjData(m)
r = mujoco.Renderer(m, 256, 256)
mujoco.mj_step(m, d)
r.update_scene(d)
img = r.render()
print(f\"EGL render OK: {img.shape}\")
\"
'"
ssh $SERVER "srun --gres=gpu:1 --time=5:00 --partition=<your-partition> bash -c '
export HF_HUB_OFFLINE=1
python -c \"import torch; ckpt=torch.load(\\\"CHECKPOINT_PATH\\\", map_location=\\\"cuda:0\\\"); print(f\\\"Checkpoint loaded: {type(ckpt)}\\\")\"
'"
Write PREFLIGHT_REPORT.md:
| Check | Status | Details |
|-------|--------|---------|
| SSH | PASS/FAIL | [hostname] |
| GPU | PASS/FAIL | [GPU model, memory] |
| Network | PASS/FAIL | [proxy status, HF reachable?] |
| Conda | PASS/FAIL | [PyTorch version, CUDA] |
| Rendering | PASS/FAIL | [EGL status] |
| Model Load | PASS/FAIL | [checkpoint type] |
ALL checks must PASS before submitting experiment jobs.
Autonomously improve a generated paper via GPT-5.4 xhigh review → implement fixes → recompile, for 2 rounds. Use when user says "改论文", "improve paper", "论文润色循环", "auto improve", or wants to iteratively polish a generated paper.
Autonomous research review loop using any OpenAI-compatible LLM API. Configure via llm-chat MCP server or environment variables. Trigger with "auto review loop llm" or "llm review".
Autonomous multi-round research review loop using MiniMax API. Use when you want to use MiniMax instead of Codex MCP for external review. Trigger with "auto review loop minimax" or "minimax review".
Autonomous multi-round research review loop. Repeatedly reviews via Codex MCP, implements fixes, and re-reviews until positive assessment or max rounds reached. Use when user says "auto review loop", "review until it passes", or wants autonomous iterative improvement.
Generate and rank research ideas given a broad direction. Use when user says "找idea", "brainstorm ideas", "generate research ideas", "what can we work on", or wants to explore a research area for publishable directions.
Workflow 1: Full idea discovery pipeline. Orchestrates research-lit → idea-creator → novelty-check → research-review to go from a broad research direction to validated, pilot-tested ideas. Use when user says "找idea全流程", "idea discovery pipeline", "从零开始找方向", or wants the complete idea exploration workflow.