skills.rest
yyccbb
yyccbbCOMMUNITY
Explore all skills by this author →
GitHub

run-visible

Launch long ML training runs in a visible terminal and monitor progress without blocking the agent.

Updated Aug 14, 2026
Search Tags:#monitoring#machine learning#bash#tmux#terminal#experiment tracking#training run
One-click install
npx skills add https://github.com/yyccbb/PaperReproduction-Skills --skill run-visible-yyccbb
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: run-visible
Source: https://github.com/yyccbb/PaperReproduction-Skills/tree/main/skills/run-visible
Command: npx skills add https://github.com/yyccbb/PaperReproduction-Skills --skill run-visible-yyccbb

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Long training or evaluation runs executed as foreground shell calls hit tool timeouts and leave the human staring at a spinner with no visibility into progress. This Skill launches runs into a terminal the human can watch live, then polls status cheaply so monitoring never floods the agent's context. ## Core Features & Use Cases - Visible Launch Backends: Detects the best available surface — tmux split, WezTerm/kitty pane, macOS Terminal/iTerm window, Linux GUI terminal, detached tmux for headless SSH, or nohup as a last resort — and prints an attach command for the user. - Structured Run Artifacts: Each run gets a timestamped directory under .paper-reproduction/runs/<RUN_ID>/ containing console.log, exit_code, and an optional status.json heartbeat for cheap polling. - Context-Safe Monitoring: Polls every ~60 seconds using status.json, a CR-translated log tail, or tmux capture-pane, reporting one line per poll instead of dumping raw logs. - Use Case: A user says "train the model with configs/a.yaml". The Skill preflights the config, spawns the run in a visible tmux pane, gives the user the attach command, and narrates step/loss/ETA until the exit code appears. ## Quick Start Ask the agent to launch your training command, for example: run python train.py --config configs/a.yaml in a terminal I can watch and keep me updated on its progress.

Frequently Asked Questions about run-visible

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a long training job without blocking my AI agent?▼

Launch the command through the spawn_visible.sh script, which starts it in a separate visible terminal and returns immediately. The agent then polls the run directory roughly every 60 seconds and reports one-line progress updates.

How do I monitor a running ML training job from the terminal?▼

Check the run's status.json heartbeat file if one exists, or tail console.log after translating carriage returns with tr. If the run is in tmux, tmux capture-pane gives the current on-screen progress bar state directly.

Does this work over SSH or on a headless server?▼

Yes. When no display is available, the launcher falls back to a detached tmux session and prints the attach command, or nohup as a last resort. Running the agent inside tmux is recommended for the best experience.

Why does tailing a training log produce one enormous line?▼

tqdm progress bars write carriage returns instead of newlines, so the whole bar history collapses into one line. Pipe the tail through tr '\r' '\n' before reading it to get normal lines.

What happens when a training run fails?▼

The launcher writes the exit code to the run directory when the command finishes. On a non-zero exit, the agent reads the last ~80 log lines, diagnoses the failure, and proposes a fix, but never relaunches without user confirmation.

Related Skills

Explore More Software Engineering Skills→
NVIDIANVIDIA
3.2k

nv-generate-mr-brain-finetune

Finetunes the NV-Generate-CTMR MR-Brain v1 diffusion UNet on NIfTI training volumes.

Official
Advanced
LightconeResearchLightconeResearch
18

figure-comparison

Compares paper reference figures, tables, and values with reproduced results.

Official
Advanced
ShinyGuaShinyGua

mm-progress-monitor

Monitor status.json and update TodoWrite checklists in real time.

Community
Intermediate
kang-jaehyunkang-jaehyun

submit

Submit machine-learning training jobs to a SLURM cluster with resource requirements.

Community
Advanced