Menu
Run a minimal train/import smoke for ML crates (Rust or Python HF). Use before claiming a training path works.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Draft or refresh a Hugging Face model card (MODEL_CARD.md / README model section). Use when publishing BitNet/trit/HF checkpoints or training runs.
Run the standard Rust local gate (fmt, clippy -D warnings, build, test). Use before commits/PRs on Rust crates.
Prepare code for git commit by running format checks, lints, tests, and generating a commit message. Use before committing changes.
Detect drift from tzervas fleet standards (workflows, badges, issue tiers, Copilot off). Use when auditing a repo or filing fleet-gap issues.
Two-pass pull request review — triage by size/risk, then grounded file:line checklist. Use when reviewing PRs or before merge.
Review Rust unsafe blocks for SAFETY comments (what/why/invariants). Use when writing or reviewing unsafe Rust.
| name | train-smoke |
| description | Run a minimal train/import smoke for ML crates (Rust or Python HF). Use before claiming a training path works. |
| metadata | {"author":"tz-forge","version":"1.0","domain":"ml"} |
| allowed-tools | Bash(cargo:*) Bash(uv:*) Bash(python:*) Bash(pytest:*) |
ml-rust-crate or ml-python-hf./scripts/check.sh --quick # or cargo test --all-features
cargo test --all-features
uv sync --all-groups 2>/dev/null || uv sync
uv run python -m {{project_snake}}.train_smoke
uv run pytest -q
wandb/, checkpoints/, *.safetensors, large .bin/.pt.Train smoke
===========
kind: ml-rust-crate | ml-python-hf
imports: ✅/❌
tests: ✅/❌
notes: ...