skills-qa

To assess a legal skill for prompt injection risks, scan its SKILL.md content for authority overrides and high-risk instruction tactics using an injection-safety heuristic. This produces a structured verdict indicating whether the skill is safe to trust or deploy.

Evaluating workflow governance in legal skills requires grading the skill against design parameters like trust surface, schema quality, and conflict detection. This framework evaluation identifies overlap conflicts and dependency breakage risks before team adoption.

Checking dependency mapping for a legal skill involves building an upstream, downstream, and auto-trigger picture to surface breakage risks. This mapping identifies incomplete context and dependencies before you install the skill into attorney workflows.

Yes, you can review pasted SKILL.md content to decide if a legal skill is trustworthy. The evaluation accepts a skill path, file path, or pasted content directly to produce a Ready or Refuse verdict for your install decision.

You need a schema validation check for legal workflow skills when reviewing first-party skills before team deployment or evaluating community skills for adoption. This check grades schema quality to ensure the skill meets design standards before installation.

The limitation of heuristic scans for prompt injection in legal skills is that they detect high-risk instruction patterns and authority overrides, but they provide a risk verdict rather than a guarantee of absolute safety. Material concerns may still require manual review.