just-vibeData engineering · inspect by default
/data-quality
Check freshness, completeness, validity, and consistency
Use to evaluate an identified snapshot against declared rules; data-profile discovers descriptive anomalies, data-pipeline embeds checks and ops-observability or ops-alerts implement monitoring.
Make it your own.
In Claude Code, use the slash command and add your context. In Codex, select data-quality from the just-vibe skill picker, then send the same brief.
Version 0.11.0 also supports /jv data-quality, /just-vibe data-quality and /jv:data-quality in Claude. See shortcut setup and context examples.
/just-vibe:data-quality Check these records against the supplied freshness and validity rules./just-vibe:data-quality Evaluate fresh-but-incomplete and complete-but-stale partitions separately./just-vibe:data-quality Assess known rules without data access; do not install monitors or invent pass rates.What the agent does
- Resolve the applicable rules and freeze their thresholds before observing results.
- Evaluate completeness, freshness and validity separately against the identified snapshot, counting excluded or unreadable records and separating warnings from failures.
- Compare with history where available and identify likely upstream causes.
Inputs
- dataset, quality contract, thresholds, time window, and bounded execution access.
Optional context: scope, references, constraints, successCriteria, environment, mode, budget.
Scope
- Reads
- Freshness, completeness, validity, uniqueness, and cross-field consistency.
- Writes
- No source changes in inspect/plan. Save only requested planning artifacts. data-pipeline embeds accepted checks, ops-observability or ops-alerts implement monitoring and data-backfill repairs records.
- Mode
- Inspect; dataset, quality contract, thresholds, time window, and bounded execution access.
- Prerequisites
- Data source/version, schema/semantics, transformation code, permitted sampling scope, and storage/compute budget. Prefer aggregates and redacted samples; never upload datasets to external services implicitly. Record time zones and snapshot identity for reproducibility.
Expected output
- Rule/version/snapshot/result matrix with affected counts, severity, likely upstream causes and repair or monitoring proposals.
How the work is checked
- Stale but complete data fails freshness; missing rule evidence is unknown rather than passing.
When to stop or clarify
- Installing monitors (data-pipeline, ops-observability, ops-alerts) or repairing records (data-backfill) is separate. Do not redefine thresholds after seeing results to force a pass.
Handling missing context
- Infer
- Inspect schema, source snapshot, transformation code, grain, time zones and permitted sample scope.
- Assume
- Use bounded synthetic or supplied samples when full data is unavailable; keep unknown values distinct from zero.
- Ask
- Resolve ambiguous entity/grain/time semantics before reconciliation or backfill; obtain missing data/compute limits only for the dependent scan or execution.
Technical guidance
- Evidence
- Resolve completeness, freshness, validity and consistency rules with denominators and consumer impact.
- Method
- Separate no data from valid zero volume; define late-arrival windows and missing-check behavior.
- Pitfall
- A green dashboard can reflect a query that stopped receiving rows rather than healthy data.
- Check
- Inject missing, stale and inconsistent synthetic batches and a healthy control; verify the right failure reason and recovery condition.
Situational decisions
When required rule evidence is missing: Mark the rule unknown and preserve the failed/unknown result rather than changing thresholds to pass.
The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.