菜单
Resolve, probe, execute, observe jobs, sync, manage service lifecycle, transfer artifacts, and clean VAWS remote Ascend session containers through structured agent-facing CLI entrypoints.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Add, verify, repair, or remove a managed remote NPU host for this workspace. Use for requests like “配置服务器”, “加一台机器”, “检查 ready”, “修容器 SSH”, or “移除机器”. Do not use for code sync, rebuilds, serving, or benchmarking.
Collect one Ascend torch-profiler case end-to-end on a workspace-managed remote NPU container. Starts a profiled vLLM service, brackets a workload with /start_profile and /stop_profile, runs analyse(), verifies kernel_details.csv landed, and writes a manifest the analysis skill can consume. Use for requests like "采集 profiling", "torch profiler 跑一个 case", "采一份 profile 出来", "采 profiling 给我分析". Do not use for pure performance benchmarking, HBM/memory profiling, or for analysing already-collected profiling data (that is the analysis skill's job).
Ensure a ready remote runtime runs the exact current local workspace state before any remote smoke, service launch, or benchmark. Use automatically immediately before remote execution when direct local -> container SSH already works and local uncommitted changes must be reflected remotely. Do not use for initial machine attach, generic Git topology work, or unrelated local-only coding.
Initialize this workspace after clone. Use for requests like “初始化仓库”, “配置 gh / GitHub 登录”, “初始化子模块”, or “把 vllm / vllm-ascend remotes 改成我的 fork”. Do not use for ordinary coding, serving, benchmarking, or unrelated Git tasks.
Start, check, or stop a single-node vLLM Ascend online service on a workspace-managed ready remote container. Use for requests like "拉服务", "在远端起个服务", "重启服务", "看服务状态", "停掉服务". Do not use for machine attach, environment bootstrap, code sync, benchmark orchestration, or offline inference.
Download, resume, status-check, and SHA256-verify ModelScope model weights. Use for $modelscope download/status/verify/check, Chinese requests to 下载/续传/补全/查看进度/校验 ModelScope 权重, and tasks that need durable background ModelScope downloads under explicit local directories.
| name | remote-toolbox |
| description | Resolve, probe, execute, observe jobs, sync, manage service lifecycle, transfer artifacts, and clean VAWS remote Ascend session containers through structured agent-facing CLI entrypoints. |
Compatibility note: .remote-dev is now the preferred local-tool-like surface
for ordinary remote endpoint development. Use this skill when the task needs
managed VAWS target resolution, session containers, parity/sync, service
adapters, artifact compatibility, or cleanup internals.
Use this skill when an agent needs to operate a managed VAWS machine or session container backend instead of writing raw SSH, scp, sftp, manual tail, or manual ps/kill commands.
--session-id or --session-file for parallel work. Use --machine
only for explicitly single-tenant legacy flows.stderr as __VAWS_REMOTE_TOOLBOX_PROGRESS__=<json>.stdout is exactly one JSON object.needs_input, blocked, failed,
timeout, needs_repair, cancelled.remote_job_start.py generate job ids by default. If you pass
--job-id, it must be globally unique in this workspace; duplicate local job
records are blocked before launching a remote process.remote_sync_plan --mode source-only and
remote_sync_apply --mode source-only must not enter install/rebuild.remote_sync_apply --mode materialize updates runtime sources but still skips
install/rebuild.remote_sync_apply --mode install delegates to remote-code-parity full
install/rebuild behavior and consent gates.Target and diagnostics:
python3 .agents/scripts/remote_target_resolve.py (--machine <alias> | --session-id <id> | --session-file <path>)
python3 .agents/scripts/remote_probe.py (--machine <alias> | --session-id <id> | --session-file <path>)
python3 .agents/scripts/remote_exec.py --session-id <id> --cwd /vllm-workspace --command 'python3 -V'
# remote_exec and remote_job_start source /etc/profile.d/vaws-ascend-env.sh by default.
# Use --no-runtime-env only when intentionally debugging the raw container shell.
Long jobs:
python3 .agents/scripts/remote_job_start.py --session-id <id> --kind build --command 'bash build.sh'
python3 .agents/scripts/remote_job_status.py --job-id <job-id>
python3 .agents/scripts/remote_job_tail.py --job-id <job-id> --lines 120
python3 .agents/scripts/remote_job_stop.py --job-id <job-id> --force
python3 .agents/scripts/remote_job_collect.py --job-id <job-id>
Sync:
python3 .agents/scripts/remote_sync_plan.py --session-id <id> --mode source-only
python3 .agents/scripts/remote_sync_apply.py --session-id <id> --mode source-only
python3 .agents/scripts/remote_sync_plan.py --session-id <id> --mode install --force-reinstall
python3 .agents/scripts/remote_sync_apply.py --session-id <id> --mode install
Service:
python3 .agents/scripts/remote_service_start.py --session-id <id> -- --model /data/models/Qwen --tp 1
python3 .agents/scripts/remote_service_status.py --session-id <id>
python3 .agents/scripts/remote_service_logs.py --session-id <id> --lines 200
python3 .agents/scripts/remote_service_stop.py --session-id <id> --force
Artifacts and cleanup:
python3 .agents/scripts/remote_artifact_manifest.py --session-id <id> --remote-path /vllm-workspace/.vaws-runtime/serving
python3 .agents/scripts/remote_artifact_pull.py --session-id <id> --remote-path /vllm-workspace/.vaws-runtime/serving --local-dir .vaws-local/remote-toolbox/artifacts/serving
python3 .agents/scripts/remote_artifact_push.py --session-id <id> --local-path ./local-report --remote-path /tmp/report
python3 .agents/scripts/remote_cleanup.py --session-id <id> --service --jobs --remote-temp --leases --dry-run
python3 .agents/scripts/remote_cleanup.py --session-id <id> --job-id <job-id> --dry-run
Local untracked state lives under .vaws-local/remote-toolbox/:
logs/ for full stdout/stderr from remote_execjobs/<job-id>.json for local job registry recordsartifacts/ for pulled job/service/profiling artifactsRemote toolbox state lives under <runtime-root>/.vaws-runtime/remote-toolbox/
inside the target container.
.agents/skills/remote-toolbox/references/behavior.md.agents/skills/remote-toolbox/references/command-recipes.md.agents/skills/remote-toolbox/references/acceptance.md.agents/skills/remote-toolbox/references/stress-validation.md