spec-validation

Run the validation script with python .claude/skills/spec-validation/scripts/validate_spec.py followed by the path to your spec.md file. It checks frontmatter fields, required sections, placeholders, and cross-references, then outputs a PASS or FAIL report.

The script verifies required YAML frontmatter fields (name, description, status, type, owner), required body sections (Overview, Scope, Acceptance Criteria), unresolved placeholders like {{TODO}} or [TBD], and whether referenced plan.md and tasks.md files exist.

Write criteria that are concrete, measurable, and verifiable by test or inspection, such as "API response time is < 200ms at p95" instead of "fast". Avoid implementation details, vague adjectives, and compound criteria; the included ACCEPTANCE_CRITERIA_GUIDE.md provides patterns and examples.

No, the skill reports findings but does not block implementation. The user decides whether the spec is ready, needs refinement, or has critical issues that should hold implementation until resolved.

Warnings appear when the spec references plan.md or tasks.md but those files do not exist yet in the spec directory, or when no reference is found. These are warnings rather than errors since plans are often created after spec approval.