eightmm

Execute auditable quantitative research across preregistered study design, statistical inference, numerical verification, dimensional analysis, uncertainty propagation, counterexample search, and formal proof. Use when a scientific or mathematical question…

Compare finite actions under uncertainty with explicit states, loss, utility or risk-neutral payoff, priors, experiment likelihoods, and a declared decision criterion. Use for Bayes risk, expected utility or payoff, pure-action minimax, minimax regret,…

Compare finite actions under uncertainty with explicit states, loss, utility or risk-neutral payoff, priors, experiment likelihoods, and a declared decision criterion. Use for Bayes risk, expected utility or payoff, pure-action minimax, minimax regret,…

Start or continue a task-scoped scientific workbench in Codex. Invoke when the user explicitly uses $codex-science or asks to start, activate, enable, load, or enter Codex Science, including "Codex Science 시작" or "Codex Science 활성화". Also invoke automatically…

Inspect and use the local computer as a reproducible scientific workbench across shell, Python, R, Julia, Jupyter, containers, CPUs, and GPUs. Use when a scientific task needs local files, code execution, package or environment setup, data conversion,…

Audit equations, models, conversions, scales, and limiting cases with explicit dimensions and units, plus executable SI dimensional-consistency and conversion receipts. Use for physical quantities, empirical correlations, nondimensionalization, scaling laws,…

Quantify and review measurement or model-input uncertainty with explicit measurands, calibration, Type A and B components, covariance, linear sensitivities, seeded Monte Carlo propagation, nonlinear disagreement, and machine-readable uncertainty receipts.

Formalize mathematical statements and produce auditable kernel-check receipts in an existing Lean 4 workspace. Use for theorem-statement audits, proof repair, definition design, tactic or term proofs, dependency minimization, axiom inspection, and…

Shared state, graph-guided work, typed runtime projections, recovery, peer consultation and review, isolated delegation, patch landing, bounded autonomy, experiments, Draft PRs, and cross-provider handoff.

Resolve implementation-shaping choices for new projects, repo onboarding, unresolved drafts, or broad features. Inspect first and ask only about material gaps; clear bounded changes do not require an interview.

Register datasets and splits in a manifest and check group leakage and split drift before training: declared group keys (not row identity) decide train/eval separation, so the same protein, patient, or source never sits on both sides.

Run training and evaluation as registered experiments instead of ad-hoc launches: check the experiment board for duplicates, pre-register hypothesis runs, gate long runs through the run ledger, capture a reproducibility capsule, and record outcomes where the…

Discipline for launching training or inference on a local GPU machine (not a Slurm cluster): check VRAM and running processes before launching, serialize GPU jobs through the tsp queue instead of racing other work, read the machine snapshot for hardware…

Work on a Slurm cluster through the private cluster reference and the harness job tools instead of re-probing the cluster every time. Use for sbatch/srun submissions, queue or partition questions, node and GPU availability, job monitoring, and post-job log…

Maintain an oh-my-setting installation from chat: status, update, doctor, duplicate-skill cleanup, local snapshots, rollback, or uninstall. Use `oms-agent-harness` instead for repository task state.

Explain why an observed failure or anomaly happened before changing anything: competing hypotheses, ranked evidence, and the cheapest probe that tells them apart. Use for regressions, silent wrong results, flaky or unreproducible behaviour, and "why did this…