Glossary
Language used across the product and documentation. For a walkthrough, see the user guide. After you sign in, the same glossary is available in the app sidebar.
- Skill
- A structured AI behavior contract: inputs, outputs, scope, guardrails, and metadata. Skills move through draft, guardrail validation, testing, and certification states.
- Workspace
- An isolated organization container. Users, skills, and invites belong to a workspace. Data does not leak across workspaces.
- Guardrail
- A policy rule (for example hallucination handling, scope enforcement, or audit expectations) attached to a skill. Guardrails must be validated before testing and certification.
- Certification
- The state after tests pass the required score and the user issues certification. Certified skills receive a certification timestamp and can be exported as artifacts.
- Test suite
- Automated evaluations of in-scope, edge, and out-of-scope behavior. A minimum score (80%) is required to proceed to certification.
- Scope / Out of scope
- Explicit lists of what the skill is allowed to do and what it must refuse. They shape guardrails and tests.
- Skill listing
- A marketplace entry that points at a certified skill, with description, ratings, and publisher information.
- Fork
- Creating a derived skill from a marketplace listing into your workspace so you can adapt it under your own governance.
- Owner / Member / Superadmin
- Owner: manages workspace members and activity. Member: standard user. Superadmin: platform administration across all workspaces.
- Data request
- A GDPR-style request for export or deletion of personal data, tracked in the app and processed according to your deployment policies.
- Export format
- JSON Schema, YAML, or platform-oriented bundles (for example Claude, Gemini) generated from a certified skill.
- Version
- Semantic version of a skill. Creating a new version starts a new draft that must complete the full pipeline again before certification.