# Agentic Prototype Critique

> Use when reviewing a generated prototype manifest and diagnostics for design completeness, contract validity, and implementation consumption readiness.

- Skill: `dolphinai2026/agentic-prototype-critique` (Agent Skill)
- Install (CLI): `npx skillmds@latest add dolphinai2026/agentic-prototype-critique`
- Raw SKILL.md: https://api.skillmd.com/api/skills/dolphinai2026/agentic-prototype-critique/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: dolphinai2026 (https://skillmd.com/u/dolphinai2026)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/dolphinai2026/agentic-prototype-critique

---


# Agentic Prototype Critique

## Trigger Conditions

Use this skill after prototype generation, before implementation consumes prototype artifacts, or when the user asks for a design critique, doctor review, or readiness assessment.

## Required Inputs

- `<workspace>` containing the generated prototype artifacts.
- `<manifest>` path to the prototype manifest.
- Active `DESIGN.md` and source spec when semantic critique is needed.

## Outputs

- Diagnostics status.
- Contract violations and blocked messages.
- Prioritized critique findings.
- Implementation consumption risks.

## Forbidden Actions

- Do not edit prototype artifacts during critique unless the user explicitly asks for refinement.
- Do not ignore doctor errors because the visual result looks acceptable.
- Do not call external runtime design generators or host-local design tools.
- Do not commit, push, reset, rebase, stash, or delete user work as part
  of this skill (`git_read_only` tier in
  `skills/agentic-git-sync/rules/git-permission-boundaries.md`); this does
  not restrict other active skills' Git contracts.

## Command Invocation

```bash
agentic-design prototype doctor --workspace <workspace> --manifest <manifest>
```

## Workflow

Run `agentic-design prototype doctor --workspace <workspace> --manifest <manifest>`, read diagnostics, compare findings against the spec and active `DESIGN.md`, then report blocking issues before advisory critique.

Classify findings into `prototype_quality` when they affect user-reviewable
visual/business quality, and `metadata_readiness` when they affect Builder,
derivation, asset-index, source hash, or CodegenPlan readiness.

## Verification Commands

```bash
agentic-design prototype doctor --workspace <workspace> --manifest <manifest>
PYTHONPATH=python .venv/bin/pytest tests/agentic_core/test_design_skills.py -q
```