JudgmentKit

• Analyze Implementation Brief — Analyze an implementation-heavy UI brief and return an activity-centered judgment packet before UI generation or styling.
• Create Activity Model Review — Create a reviewable activity model candidate from a UI brief, with guardrails and targeted questions before UI generation.
• Recommend Surface Types — Recommend a purpose-based UI surface type from activity evidence before workflow review or frontend implementation, including artifact_inspector only when its mandatory evidence is satisfied.
• Recommend Ui Workflow Profiles — Recommend optional UI workflow guidance profiles, including artifact-inspector-ui when its artifact activity evidence applies, without applying them automatically.
• Review Activity Model Candidate — Review an inferred or externally proposed activity case against the source brief, attributed context, provenance, and JudgmentKit guardrails.
• Review Ui Workflow Candidate — Review an externally proposed UI workflow candidate against the source brief, activity review, and JudgmentKit guardrails.
• Review Cognitive Dimensions Candidate — Review a workflow or implementation candidate with Cognitive Dimensions checks for mapping, visibility, hidden dependencies, commitment, progressive evaluation, change cost, mental operations, and disclosure.
• Create Ui Implementation Contract — Create an implementation contract for generated UI, using JudgmentKit design-system authority by default, optional repo-local component authority, or a complete external design-system adapter when supplied. surface_type artifact_inspector activates scoped JudgmentKit chrome and overlay authority while keeping the primary artifact external_not_reviewed.
• Preflight Ui Implementation Candidate — Validate an implementation evidence packet and declared rendered selectors before substantive review. Returns precise field/selector repairs without consuming an implementation attempt. Admission does not prove UI quality, source authority, or product acceptance.
• Review Ui Implementation Candidate — Review generated UI or code evidence against an active UI implementation contract. The active design-system gate is hard: missing or failing design_system_provenance means the candidate is not an artifact and must be repaired and resubmitted. Include local_component_authority_evidence when repo-local component authority is active. When visual_composition_policy applies, the hosted MCP renders exact self-contained HTML in an isolated browser, measures declared and discovered relationships, and at
• Create Ui Generation Handoff — Create a UI generation handoff from a ready UI workflow review packet and a UI implementation contract, blocking non-ready reviews.
• Create Frontend Generation Context — Create adapter-layer frontend implementation context from a ready UI generation handoff, resupplying the current raw activity brief and exact attributed context so protected risk and workflow authority can be independently revalidated.
• Create Frontend Implementation Skill Context — Create gated frontend implementation skill context from the current raw activity brief, its attributed context, and a ready frontend generation context. The raw source is revalidated before implementation guidance is emitted.
• Create Slide Deck — Plan or create a JudgmentKit PPTX slide deck from user-facing slide content using the presentation-theme template registry. Without output.path or dry_run false, this returns a safe deck plan. Artifact creation requires a local @oai/artifact-tool runtime and writes only under outputs/judgmentkit-slide-decks.
• List Icon Catalog — Page through the committed Lucide icon catalog metadata. Set include_svg true only when a page needs SVG-ready data.
• Search Icon Catalog — Search the committed Lucide icon catalog by id, name, alias, tag, or tokenized terms and return ranked matches.
• Get Icon Svg — Return the full generated Lucide icon record and inline SVG for one canonical icon id.

Schemas and invoke metadata: Directory API