Menu
Checklist before long execution_run or execution_run_background for ML/training scripts.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
MaxEvolve Triton/PyTorch to JAX/Pallas porting with GpuToJax 12-step pipeline, interpret=True validation, MAP-Elites evolution, and Colab/SSH hardware profiling.
Manage Google Colab sessions and execute code on remote Colab VMs via the colab-cli.
Use when you need to generate a synthetic dataset. You can use Afterimage Python package to create, conversational, tool-calling, multi-choice questions, or other types of structured output datasets for SFT, or preference data for DPO.
Schedule deferred or recurring work via the isanagent cron tool—add/remove jobs with correct runtime context fields.
Configure long wall clocks and use execution_run_background + job polling for ML or hours-long runs without blocking the agent loop.
Run numeric or ML code through isanagent execution tools without flooding context—artifacts, incremental runs, sandbox paths.
Based on SOC occupation classification
| name | ml-execution-preflight |
| description | Checklist before long execution_run or execution_run_background for ML/training scripts. |
| always | true |
Run these checks before a long or GPU-heavy execution_run / execution_run_background:
execution_env_info — note max_wall_secs, default timeout, provider id, and artifact caps.execution_run that prints Python version and (if relevant) torch.cuda.is_available() or equivalent.timeout_secs explicitly from model size and hardware; training almost always needs a large value within max_wall_secs.execution_artifact_list, sandbox paths, or copy-out plan). Ephemeral sandboxes lose data when torn down.If any step fails, fix the root cause before scaling.
**Before launching a fine-tuning or dataset generation script, also load long-running-execution with load_skill_instructions.