Orchestrate an end-to-end data analysis or machine learning project from decision framing through reproducible handoff. Use when a request spans multiple lifecycle stages or an ambiguous modeling request must become a decision-ready result; use narrower audit or experiment-design skills for isolated reviews.
The English summary is being prepared. Showing the original description.
See how to installReview the instructions the agent will receive before you install.
Produce work that another person can inspect, challenge, rerun, and use for a real decision. Treat a model score as evidence, not as the deliverable.
Before choosing methods, make the following explicit:
Infer reversible details when evidence supports them. Ask for a decision only when alternatives would materially change the target, evaluation, permitted data, or external impact. Record assumptions rather than repeatedly pausing for routine choices.
For a full gate-by-gate procedure and stopping conditions, read references/project-gates.md.
auditing-data-and-ground-truth skill for a dedicated investigation; otherwise perform the checks directly and keep their evidence in this project's record.designing-leakage-safe-experiments skill for nontrivial feature availability, group dependence, or temporal leakage; otherwise document the same boundaries in the evaluation protocol.diagnosing-ml-failures skill when installed; otherwise isolate the first divergent layer and retain a competing-hypothesis log before proposing a fix.validating-models-and-claims skill when installed; otherwise map each falsifiable claim to independent, scope-matched evidence and narrow unsupported wording.shipping-reproducible-results skill when installed; otherwise attempt a clean independent run and label the result as reproduced, ready for independent reproduction, replayable with constraints, or archival evidence only. Never imply that an unexecuted path was reproduced.Continue autonomously through ordinary, reversible analysis. Obtain approval before changing the target or source of truth, discarding or overwriting original data, adopting a consequential metric tradeoff not already authorized, incurring material cost, or publishing/deploying externally.
Never treat approval as proof. Keep the evidence, rationale, owner, and downstream effect of each material choice.
Copy assets/decision-grade-project-record.md into the project and adapt it. Keep claims linked to evidence, not merely to code or charts. Clearly label each final claim as supported, conditional, contradicted, or not tested, and track reproduction status separately.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Audit datasets, joins, labels, and ground truth before analysis or modeling. Use when data meaning, row grain, time semantics, source-of-truth reliability, or label construction may invalidate conclusions; not for general model evaluation after the evidence base is already trusted.
The English summary is being prepared. Showing the original description.
Design leakage-safe machine learning experiments that mirror real deployment and support fair model comparisons. Use when defining prediction timing, feature eligibility, train-validation-test splits, baselines, metrics, or controlled model iterations; not for auditing whether raw labels are trustworthy.
The English summary is being prepared. Showing the original description.
Isolate the root cause of ML performance drops, inconsistent evaluations, prediction errors, and training-serving mismatches across data, labels, splits, pipelines, models, metrics, and runtime behavior. Use when investigating a reproducible failure or regression, not routine model selection or general performance validation.
The English summary is being prepared. Showing the original description.
Package completed data analysis and ML work so an independent recipient can reproduce the claimed results, verify artifact lineage, and operate the handoff within its stated scope. Use when finalizing a project, study, model package, or review bundle; not for deploying to a live system.
The English summary is being prepared. Showing the original description.
Route data analysis and machine learning work to the right skill in this suite. Use when starting any analysis, modeling, validation, or reproducibility task and the matching specialized skill is not yet clear; not needed when one specific skill already clearly applies.
The English summary is being prepared. Showing the original description.
Validate trained models and analytical claims against their intended decision, independent evidence, and human-reviewed ground truth. Use when reviewing model performance, analysis conclusions, launch claims, or evaluation reports; use failure diagnosis instead when the main task is locating a known defect.
The English summary is being prepared. Showing the original description.