# Praxis Visual Builder AI Validation

> Praxis Visual Builder AI Validation

- Skill: `codexrodrigues/praxis-visual-builder-ai-validation` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx get-skill add codexrodrigues/praxis-visual-builder-ai-validation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/codexrodrigues/praxis-visual-builder-ai-validation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: codexrodrigues (https://skillmd.com/u/codexrodrigues)
- Updated: 2026-11-15
- Page: https://skillmd.com/skills/codexrodrigues/praxis-visual-builder-ai-validation

---


# Praxis Visual Builder AI Validation

Use this skill for governed AI authoring of visual-builder rules. AI operations must resolve semantic intent into declared manifest operations and validators; they must not patch arbitrary JSON or route primary intent through keywords.

Pair it with:

- `praxis-ai-authoring-manifests` for component manifest structure, operation contracts, targets, validators, and examples.
- `praxis-ai-registry-ingestion` when manifest changes must be discovered by the AI registry.
- `praxis-ai-semantic-intent` when user intent resolution, grounding, or tool selection is involved.
- `praxis-visual-builder-graph-runtime` for graph state, node identity, edges, validation errors, and registry integrity.
- `praxis-visual-builder-jsonlogic-roundtrip` for `condition.set`, JSON Logic parsing, and `dsl.roundTrip.validate`.
- `praxis-visual-builder-schemas-templates` for `variable.add`, `effect.set`, target schemas, field schemas, and context variables.
- `praxis-page-builder-ai-agentic` when visual-builder AI operations are delegated from Page Builder agentic authoring or child widget config flows.

## Source Audit

Before editing visual-builder AI behavior, inspect:

- `projects/praxis-visual-builder/AGENTS.md`
- `projects/praxis-visual-builder/README.md`
- `projects/praxis-visual-builder/src/lib/ai/praxis-visual-builder-authoring-manifest.ts`
- `projects/praxis-visual-builder/src/lib/ai/visual-builder-ai.adapter.ts`
- `projects/praxis-visual-builder/src/lib/ai/visual-builder-ai-capabilities.ts`
- `projects/praxis-visual-builder/src/lib/ai/visual-builder-ai.context.ts`
- `projects/praxis-visual-builder/src/lib/ai/visual-builder-context-pack.ts`
- `projects/praxis-visual-builder/src/lib/services/rule-builder.service.ts`
- `projects/praxis-visual-builder/src/lib/services/rule-node-registry.service.ts`
- `projects/praxis-visual-builder/src/lib/services/context/context-management.service.ts`
- manifest, adapter, AI wizard, and AI integration specs

Also inspect core AI manifest/registry contracts if operation schema, validator shape, or ingestion behavior changes.

## Canonical AI Boundary

`PRAXIS_VISUAL_BUILDER_AUTHORING_MANIFEST` owns the agentic contract for:

- editable targets: `node`, `edge`, `variable`, `condition`, `effect`, and `dslDocument`;
- operations: `node.add`, `node.remove`, `node.configure`, `edge.connect`, `edge.remove`, `variable.add`, `condition.set`, `effect.set`, and `dsl.roundTrip.validate`;
- input schemas, identity keys, affected paths, preconditions, destructive flags, failure modes, validators, and examples;
- round-trip requirements for graph identity, JSON Logic, target property schemas, and child component ownership.

`VisualBuilderAiAdapter` owns guarded `RuleBuilderConfig` patch application for the component. It must not become a bypass around graph operations, registry integrity, JSON Logic validation, or target-property validation for persisted rules.

## AI Authoring Rules

- Resolve user intent semantically into manifest operations. Do not decide intent through keyword lists, regexes, or local aliases.
- Use matching, labels, field search, and fuzzy ranking only after the semantic operation scope and canonical target kind are known.
- Every AI operation needs stable identity keys: `nodeId`, source/target node IDs, variable scope/name, effect target type/targets, or round-trip format/compare mode.
- Destructive `node.remove` and `edge.remove` require explicit confirmation when semantics can be lost.
- `condition.set` must provide JSON Logic objects within the supported visual subset, not JavaScript strings.
- `effect.set` must validate `targetType`, target IDs, governed properties, and property value types through target property schemas and `RULE_PROPERTY_SCHEMA`.
- `variable.add` governs contextual/template flows; visual field-condition references to context variables remain unsupported until the canonical JSON Logic shape is defined.
- Host-provided widget/page context must enter through declared context packs, field schemas, target schemas, or variables. Do not let Page Builder or another host inject arbitrary prompt context that becomes persisted rule state.
- `dsl.roundTrip.validate` must prove export/import stability and reject semantic drift or unstable identity.

## Inventory Before New Contract

Classify AI gaps before adding manifest operations:

- `ja-suportado-so-ux`: the operation, validator, failure mode, or example exists but the assistant/wizard does not expose it well.
- `ja-suportado-mal-nomeado-ou-mal-materializado`: broad config patches or host-local labels are being used where manifest targets and operation IDs already exist.
- `suportado-parcialmente`: the manifest operation exists but handler wiring, validator coverage, adapter behavior, examples, or registry ingestion is incomplete.
- `lacuna-real-de-contrato`: no declared target, operation, input schema, validator, handler contract, or canonical runtime path can express the AI authoring decision.

Only a real gap justifies new AI operations or public authoring contracts. When adding one, define source of truth, affected paths, failure modes, examples, registry ingestion impact, docs/examples, and local validation.

## Validation

Use focused local gates:

- `praxis-visual-builder-authoring-manifest.spec.ts` for manifest shape, operations, validators, examples, and round-trip requirements;
- `visual-builder-ai.adapter.spec.ts` for patch application, snapshot/restore, and failure behavior;
- AI wizard/integration specs when user-facing AI authoring changes;
- graph, JSON Logic, schema/template skills for operation-specific runtime validation;
- host delegation specs when AI operations are invoked through Page Builder or another component authoring surface;
- `praxis-ai-registry-ingestion` checks when generated registry assets or catalog ingestion changes;
- `npm run build:praxis-visual-builder` when public AI exports change.

Final audit should state whether manifest validation, adapter behavior, operation runtime effects, registry ingestion, and JSON Logic round-trip were tested locally.

