"Routes safe command-generation and launcher preset workflows for
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill experiment-launchers --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Experiment Launchers?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-experiment-launchers)More formats (shields.io, HTML) on the badges page.
---
name: experiment-launchers
description: "Routes safe command-generation and launcher preset workflows for
CUT, FastCUT, SinCUT, and pretrained experiments."
disable-model-invocation: true
metadata:
disco-role: operating
license: NOASSERTION
---
# experiment-launchers
Use this sub-skill when the user asks about `python -m experiments`, launcher presets, or how to list the concrete training/testing command strings for the repository's known experiment families.
## Read this when
- The user wants the preset commands for `grumpifycat`, `pretrained`, or `singleimage`.
- The user wants to inspect commands without starting tmux panes or running training.
- The user asks what the launcher `id` argument means.
- The user hits the broken `dry` path in this checkout.
- The user wants to map launcher output back to `train.py` or `test.py` commands.
## What this sub-skill covers
- Launcher CLI shape: `python -m experiments <name> <cmd> <id...>`.
- Supported families: `grumpifycat`, `pretrained`, and `singleimage`.
- Supported actions such as `run`, `train`, `test`, `run_test`, `print_names`, and `print_test_names`.
- GPU selection and epoch continuation behavior in the launcher wrapper.
- A safe bundled command-list helper that does not call tmux or execute training.
## What this sub-skill does not cover
- Actual training or testing execution. Read `../translation-workflows/` for that.
- Dataset conversion or layout preparation. Read `../data-preparation/` for that.
- `placeholder_launcher.py` as a supported workflow; it references stale or unsupported options.
## Use the bundled helper
Run `scripts/list_experiment_commands.py` to inspect known presets safely:
```bash
python scripts/list_experiment_commands.py --family grumpifycat --kind train
python scripts/list_experiment_commands.py --family pretrained --kind test --ids 2,3
python scripts/list_experiment_commands.py --family singleimage --kind both --json
```
The helper prints command strings only; it never starts tmux, allocates GPUs, downloads datasets, or runs model training.
## Read next
- `references/launcher-reference.md` for preset command tables and CLI behavior.
- `references/troubleshooting.md` for the broken `dry` command, missing `id`, and tmux/GPUtil issues.
- `../translation-workflows/references/cli-reference.md` when you need to edit the commands the launchers emit.
## Quick reminders
- `print_names` and `print_test_names` were verified as safe inspection commands.
- The upstream `dry` command fails in this checkout because its method call omits required ids.
- `launch`, `launch_test`, `stop`, and `close` can create or kill tmux windows; do not use them merely to inspect command strings.

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!