Validating Context Completeness

Quick start

Collect or infer:

• •Context pack to validate (full JSON/YAML or markdown context file)
• •Feature scope (single feature or multi-feature system)
• •Target consumer (human writer, AI agent, or both)

Then produce output using TEMPLATES.md. Validate with RUBRIC.md.

Workflow

1. •Run structural validation: check all required sections exist.
2. •Run semantic validation: check states reference valid transitions, vocabulary is used consistently, no orphan definitions.
3. •Run coverage validation: ensure every state has error handling, every action has success/failure paths.
4. •Generate validation report with severity levels (blocking, warning, suggestion).
5. •If blocking issues exist, document remediation steps.
6. •Run the rubric check. Revise until it passes.

Degrees of freedom

Level: Low

• •Default: follow validation rules exactly
• •Allowed variation: severity thresholds may adjust for draft vs production context packs

Behavioral logic

Context packs fail silently when incomplete. A missing state definition causes AI to hallucinate. A conflicting term causes user confusion. This skill enforces completeness as a gate, not a suggestion.

Validation must answer:

• •Can every state be reached from another state?
• •Does every error have a recovery path defined?
• •Are all vocabulary terms used in the pack defined?
• •Do any definitions contradict each other?

Decision framework

References

• •Templates: TEMPLATES.md
• •Rubric: RUBRIC.md
• •Examples: EXAMPLES.md
• •Required fields: reference/required-fields.md
• •Validation rules: reference/validation-rules.md
• •Scripts: scripts/