remote-execution-cli
Configure ClearML and build safe remote execution commands or SDK launch flows with clearml-task, clearml-init, clearml-debug, Task.create, Task.clone, Task.enqueue, and Task.execute_remotely.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- remote-execution-cli
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Configure ClearML and build safe remote execution commands or SDK launch flows with clearml-task, clearml-init, clearml-debug, Task.create, Task.clone, Task.enqueue, and Task.execute_remotely.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/cli-reference.md7.0 KB
- 参考文档references/troubleshooting.md7.1 KB
- 参考文档references/workflows.md7.1 KB
03 SKILL.md 正文
ClearML Remote Execution CLI
Use this sub-skill when an agent needs to configure ClearML, launch code on a clearml-agent, create or clone a remote task, enqueue an existing task, import an offline session, or diagnose why a clearml-task/remote execution request is not running.
Start Here
- Confirm the user has a ClearML server target and credentials configured via environment variables or a ClearML config file; use
scripts/validate_clearml_environment.pyfor a read-only signal check that does not print secret values. - Decide whether the launch is CLI-first (
clearml-task) or SDK-first (Task.create,Task.clone,Task.enqueue,Task.execute_remotely). Use CLI-first for CI/CD and command construction; use SDK-first when code already owns aTaskobject or needs programmatic clone/enqueue behavior. - Build commands with
scripts/build_clearml_task_command.py, which validates incompatible options and prints a shell-safe command without contacting a ClearML server. - Use
references/cli-reference.mdfor command flags and templates,references/workflows.mdfor launch patterns, andreferences/troubleshooting.mdfor failure triage.
Route Related Work
- For metrics, plots, logs, models, or artifacts inside the launched script, use
../experiment-tracking/SKILL.md. - For
clearml-data,Dataset,StorageManager, or dataset storage workflows, use../data-storage/SKILL.md. - For pipelines, pipeline CLI flags, schedulers, services, or HPO, use
../automation-pipelines/SKILL.md.
Safe Operating Rules
- Do not run
clearml-task,clearml-init,clearml-debug token, or SDK launch/enqueue code unless the user explicitly wants side effects against a ClearML server. - Prefer producing a reviewed command and validation checklist; remote execution requires credentials, network access, and a polling
clearml-agenton the selected queue. - Never echo ClearML secret values in instructions, logs, or generated scripts. Redact or report presence only.
- Treat repository, Docker image, package, requirements, queue, and output URI as deployment decisions; ask when any are missing or ambiguous.
- If no
--queueis provided,clearml-taskcreates a draft task but does not launch it.
Minimum Launch Checklist
--projectand--nameare set unless using--import-offline-session.- Exactly one entry style is chosen:
--script,--module,--base-task-id, or--import-offline-session. - Code source is explicit:
--repofor a remote repository,--folderfor a local git checkout, or--skip-repo-detectionwhen source capture is intentionally disabled. - Environment is explicit enough for the agent:
--requirements,--packages,--force-no-requirements, or--skip-python-env-install. - Queue name matches a live ClearML agent pool; otherwise the task remains queued.
- Use
--skip-task-initonly when the script already callsTask.init()or should not be patched byclearml-task.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.