Critique To Dex

Critique → Dex

Turn a prioritized critique into an actionable Dex epic. Read a findings report, normalize it to the priority findings contract, then create a single epic with one task per actionable finding — ranked by priority, each carrying enough context (what/why/how/done-when) to be worked without the original report.

This skill is source-agnostic. It works cleanly with ui-design-critique output and with any future critique that emits the contract in references/priority-findings-contract.md.

For dex command semantics and the repo-local storage guardrail, this skill relies on the dex skill — do not duplicate that knowledge here; apply it.

Inputs

Accept the findings from any of:

Process

1. Locate and normalize the findings

Extract every finding into the contract (references/priority-findings-contract.md): subject, and per finding id, title, priority, classification, location, issue, impact, evidence, recommendation.

• Normalize priority to a rank using the table below, whatever scheme the source used.
• If a required field is missing and cannot be inferred from context, note it; do not invent evidence.

2. Verify Dex storage scope

Before creating anything, confirm Dex is writing to repo-local storage (run dex dir; if it resolves to a global path, fix the workspace first). Follow the dex skill's repo-local storage guardrail exactly. Never create a project epic in global Dex by accident.

3. Build the epic + task plan

Apply the mapping rules below to produce:

• one epic for the whole critique,
• one task per actionable finding,
• optional subtasks when a single finding has clearly separable parts,
• optional blockers when one fix must precede another (only if the source states or clearly implies a dependency).

4. Preview and confirm

Creating Dex tasks writes files. ALWAYS show the plan first and get confirmation before creating:

## Dex plan preview — [subject]
Epic: [epic title]  (strengths preserved in epic description; N strengths)
Tasks (M):
| # | Priority → dex | Class | Title |
|---|----------------|-------|-------|
| 1 | P0 → 1 | Gap | ... |
| 2 | P1 → 2 | Inconsistency | ... |
Excluded: [e.g. 3 Suggestions (P3) — include? / Strengths → not tasks]

Ask once: include Suggestions as tasks, or list-only? Default: include P0–P2 findings of any class; include P3/Suggestions only on confirmation.

5. Create in Dex

Create the epic first, then tasks in priority order. Build each task description with the "good context" shape (see mapping). Attach --priority and --parent/--blocked-by as planned. Capture the returned IDs.

6. Report

Return the epic ID, the ordered task IDs with their priorities, what was excluded, and the strengths preserved in the epic description.

Mapping Rules

Priority normalization

(dex priority: lower number = higher priority. If a source uses a 1–5 or reversed scheme, map by ordinal position, highest-impact → 1.)

Classification handling

Epic

• Title: [Critique type]: [subject] — e.g. Design critique: Checkout page.
• Description: the report's verdict/summary, the finding counts by priority, a "Preserve — do not regress" list built from Strengths, and a pointer to the full report (file path if one exists).

Task description (per finding)

Build the "what / why / how / done-when" shape Dex expects:

## What
[issue] — at [location]

## Why
[impact]

## How
[recommendation]
Evidence: [evidence]

## Done when
[acceptance derived from the recommendation — the observable state that means it's fixed]

Keep the source finding id (e.g. F-01) in the task title suffix so tasks trace back to the report: Fix unreadable CTA contrast (F-03).

Output Format

## Created Dex epic — [subject]
- **Epic:** <epic_id> — [title]
- **Tasks:** N created
  - <id> · P0→1 · Gap · [title]
  - <id> · P1→2 · Inconsistency · [title]
  - ...
- **Excluded:** [suggestions/strengths and why]
- **Preserved in epic:** [count] strengths
- **Next:** `dex start <first_id>`

Reference Files