task-progress-stream

Use this skill when the user wants real-time progress updates inside the OpenClaw chat UI for a long-running task such as:

• - model training
• evaluation
• inference
• data preprocessing
• long shell jobs
• benchmark runs

It works in two modes:

1. 1. run

1. 2. tail

What it extracts

It tries to recognize common fields from logs:

• - epoch / maxepoch
• step / maxstep
• loss
• learning rate
• validation metric
• best metric
• ETA
• speed

Typical usage

Start a training job and stream progress into chat

```bash
node ./scripts/taskprogressstream.js run \
--session main \
--label train \
--cwd /path/to/project \
--cmd "python src/train.py --config configs/train.yaml" \
--interval-sec 20