Source command megaminx status

Skill Erlemar/cayley-puzzles/.agents/skills/source-command-megaminx-status

Neural distance heuristics + GPU/TPU beam search for the CayleyPy IHES Picture Cube and Megaminx puzzles

Install

npx -y skills add Erlemar/cayley-puzzles --skill source-command-megaminx-status

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

One-shot status across GCP solves, Kaggle kernels, and local runs for the megaminx competition.

SKILL.md

3.1 KB, as published. Nobody here has run it

source-command-megaminx-status

Use this skill when the user asks to run the migrated source command megaminx-status.

Command Template

/megaminx-status

Bundles the status checks we run repeatedly during long solves. Reports compact progress for each active workstream so you can decide what to do next.

Execution

Do the following in parallel where possible (one Bash call per workstream):

1. GCP VM (Phase 1, Phase 2, or any cayley-gpu run)

IP=<GCP_VM_IP>  # or read from <SSH_KEY> if set
ssh -i <SSH_KEY> \
    -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null \
    <user>@$IP \
    "ls -la /home/<user>/cayley/megaminx/submissions/ 2>&1 | tail -8; \
     echo '---active solves---'; \
     ps -ef | grep -E 'python.*03_solve' | grep -v grep | head -3; \
     echo '---tail of newest log---'; \
     ls -t /home/<user>/cayley/megaminx/submissions/*.log 2>/dev/null | head -1 | \
       xargs -I{} bash -c 'echo \"-- {} --\"; tail -8 {}'; \
     echo '---chain log---'; \
     cat /home/<user>/run_phase2_chain.log 2>&1 | tail -10"

Report: which scripts running, latest progress line (pid=N attempts=N/M), any chain status.

2. Local 4090 (Phase B or any local solve)

ls -la megaminx/submissions/*.log 2>&1 | tail -5
echo '---active processes---'
tasklist 2>&1 | grep -i python | head -5
echo '---GPU utilization---'
nvidia-smi --query-gpu=memory.used,memory.total,utilization.gpu --format=csv 2>&1 | head -2
echo '---tail of newest log---'
ls -t megaminx/submissions/*.log 2>/dev/null | head -1 | xargs tail -8

Report: GPU util, memory, latest progress line.

3. Kaggle (training kernels and probes)

export KAGGLE_API_TOKEN=$KAGGLE_API_TOKEN
.venv/Scripts/kaggle.exe kernels list --user artgor --mine -p 1 -s 5 2>&1 | head -20

For any kernel showing running, also fetch:

.venv/Scripts/kaggle.exe kernels status artgor/<slug> 2>&1

Report: which kernels running / done / cancelled, last-completed time.

Output format

Print a compact 3-section block:

GCP (<GCP_VM_IP>):
  - <script>: pid=N attempts=N/M (M-N remaining, ETA Xh)
  - chain: <last status line>

Local (4090):
  - <script>: pid=N attempts=N/M (GPU XX% util, X MiB)

Kaggle:
  - <slug>: <status> (started Xh ago)

End with a one-line "next obvious action" suggestion (e.g. "wait Xh for Phase B", "submit current Phase 1 csv", "Kaggle kernel needs ckpt download").

Gotchas

kaggle kernels output while RUNNING returns nothing — only status works mid-run.
After a kernel finishes (even cancelled/killed), output returns the partial /kaggle/working/ contents.
GCP IP changes when the VM is stopped+started — re-fetch from gcloud compute instances list if SSH fails.
03_solve.py exits non-zero at end of full-range solves due to a benign UnboundLocalError on report (fixed in main, but pre-fix runs still affected). If you see exit code 1 on a finished solve, check the CSV row count, not the exit code.