/validation-patterns

internal
Category:
Internal Skills

What it is

Reusable validation building blocks used by the validator -- factual claim extraction, filesystem verification, cross-referencing against ground truth and related artifacts, severity classification (CRITICAL/WARNING/INFO), and findings formatting.

Called by

What it does

Defines what constitutes a verifiable factual claim: file references, version numbers, config values, component names, counts, endpoints, dependencies, and ADR references.

Provides the filesystem verification method: normalize paths, check existence via Glob, verify content via Read, skip binary files.

Defines the severity classification decision tree: CRITICAL for missing files, WARNING for content mismatches, INFO for out-of-boundary references.

Formats findings into the standardized report structure consumed by all validation commands.

When you will see it surface

  • You see this skill's output every time you run a validation command -- the findings table with severity, location, and evidence columns follows the patterns defined here.

Technical notes

  • Only verifiable claims become findings -- opinions, goals, and aspirations are ignored.
  • Out-of-boundary paths (references outside the project root) are INFO severity, not WARNING.
  • Missing ground truth is not a failure -- filesystem verification runs independently.