runboard

name	runboard
description	Open a Herdr tail tab for a harness run when the user explicitly asks for a runboard, tail, Herdr tab, or raw log view. For new harness/run_batch.py monitoring, prefer the structured dashboard at scripts/run_dashboard.py reading results/_runs/<run_id>; this skill is the compatibility/tail workflow.

runboard

A runboard is a live .out file plus a Herdr tab tailing it. For new batch launches, prefer the structured dashboard:

python3 scripts/run_dashboard.py --state-root results/_runs --port 8765

Use this skill when the user specifically wants a Herdr/tail view or when a legacy run only has runs/<run>/track.out. Prefer the helper; do not hand-roll Herdr discovery unless the helper fails.

Native line shape:

[71/113] onedump-dump-encryption-pipeline / observational-memory / rep0  ok

Status legend: ok | empty | timeout | transient | exit=<n>.

Tight path: open an existing live track

Use this path when runs/<run>/track.out already exists, especially for legacy harness/run_batch.py ... | tee runs/<run>/track.out launches that predate structured state.

1. Run the helper:

   scripts/open_runboard.py --run '<run>' --label '<short label>'
   # or: scripts/open_runboard.py --track runs/<run>/track.out --label '<short label>'
   

2. Report only the returned pane, track, reused, and current progress from visible_preview.

Completion: helper output has "status": "live". The helper has already proved Herdr env, current workspace, exact tail -f process, and visible runboard content.

If there is no track.out

Only for historical/result-only runs, build a compatibility tracker first. Read result-tracker.md, then rerun the tight path.

Completion: runs/<run>/track.out exists and the helper returns status=live.

Guardrails

• If the helper says HERDR_ENV != 1, stop; do not control Herdr from outside Herdr.
• Do not use runboard as the default monitor for a new run if structured results/_runs/<run_id>/ state is available; use the dashboard instead.
• Do not dump panes, containers, token counts, streamed text, or repeated state snapshots. A runboard is one native line per cell.
• Do not claim success from a pidfile or existing track.out; success requires the helper's verified status=live result.
• For single harness/run.py cells, there is no batch runboard; tail that cell's logs only if asked.