just-vibeML data · inspect by default
/ml-dataset
Audit whether data can support the modeling task
Use to assess whether data supports a task; data-profile summarizes its columns.
Make it your own.
In Claude Code, use the slash command and add your context. In Codex, select ml-dataset from the just-vibe skill picker, then send the same brief.
Version 0.11.0 also supports /jv ml-dataset, /just-vibe ml-dataset and /jv:ml-dataset in Claude. See shortcut setup and context examples.
/just-vibe:ml-dataset Assess whether this dataset covers the intended deployment population./just-vibe:ml-dataset Assess a dataset with many rows but no labels for recently enrolled users./just-vibe:ml-dataset Review dataset metadata without row access; avoid inferring representative coverage.What the agent does
- Compare the collection and selection process and follow-up windows with the task's needs and the deployment population.
- Inspect coverage by cohort and time, identify censored or missing outcomes and missing-label patterns, and name deployment populations the data does not cover.
Inputs
- framed task, dataset version, collection process, and sampling budget.
Optional context: scope, references, constraints, successCriteria, environment, mode, budget.
Scope
- Reads
- Task suitability, population coverage, missingness, dependencies, and collection bias.
- Writes
- No source changes in inspect/plan. Save only requested planning artifacts. A separately requested repair uses the relevant implementation workflow.
- Mode
- Inspect; framed task, dataset version, collection process, and sampling budget.
- Prerequisites
- Task definition, dataset identity, field semantics, entity/time keys, and permission to inspect bounded data. Record prediction moment, label horizon, sampling, and provenance. Preserve held-out evaluation boundaries; no data upload, label alteration, or feature fitting across splits implicitly.
Expected output
- Dataset readiness report with a population/coverage table, missingness and selection risks, exclusions, supported deployment claims and needed collection or validation work.
How the work is checked
- A dataset missing outcome follow-up is not treated as fully labeled; a deployment cohort absent from training is flagged.
When to stop or clarify
- Do not infer representativeness from sample size alone. Unsupported collection semantics remain unknown.
Handling missing context
- Infer
- Read prediction moment, label horizon, entity/time keys, split policy and dataset provenance from the task and manifests.
- Assume
- Use explicit synthetic examples for design when raw data is unavailable; do not infer missing labels or fit preprocessing across held-out boundaries.
- Ask
- Ask when unresolved label timing, grouping or target semantics would change the split/features; do not demand a full dataset to explain the method.
Technical guidance
- Evidence
- Inspect collection mechanism, row grain, entity coverage, duplicates, labels and permission to use the data.
- Method
- Assess whether observations cover the intended deployment population and whether outcomes are observable without selection artifacts.
- Pitfall
- More rows do not remove survivor bias or dependence between observations.
- Check
- Reconcile sample membership and label coverage by important groups; distinguish observed gaps from unsupported population conclusions.
Situational decisions
When training data excludes the intended deployment cohort: Limit generalization claims and propose evidence collection before model complexity.
The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.