# Recap Validate

> Use when the user types /recap validate, asks Recap Studio to "score the page", "run validation", "qa the recap", or wants the multi-agent quality report regenerated. Dispatches the validation board and writes a structured report.

- Skill: `aboudjem/recap-validate` (Agent Skill)
- Install (CLI): `npx skillmds@latest add aboudjem/recap-validate`
- Raw SKILL.md: https://api.skillmd.com/api/skills/aboudjem/recap-validate/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: aboudjem (https://skillmd.com/u/aboudjem)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/aboudjem/recap-validate

---


# /recap validate

You run the validation board and emit a structured quality report.

## Inputs

- A `RecapPageContent` JSON (loaded via
  `packages/content-pipeline/src/load.ts`).
- The rendered Next.js page output (built tree under `apps/recap-web/.next/`)
  if available; otherwise reviewers work from the JSON only and downgrade
  performance & a11y confidence accordingly.

## Reviewers (parallel, isolated)

1. `fact-checker`: every claim must reference a `sourceMap` id and the
   source must support the claim. Unsupported claims drop the score.
2. `beginner-reviewer`: read time ≤ 5 min for a smart 18-year-old.
3. `accessibility-reviewer`: chunking, motion, focus, density.
4. `ux-design-reviewer`: hierarchy, polish, storytelling, responsiveness.
5. `performance-reviewer`: bundle, lazy-load, static-first, CWV risk.
6. `security-privacy-reviewer`: secrets, PII, prompt-injection cues,
   side-effect surfaces, untrusted-source flags.
7. `skeptical-reviewer`: overclaim, marketing fluff, hidden complexity.

## Scoring

Each reviewer emits `{ score: 1..10, findings: Finding[], blockers: string[] }`.

## Thresholds (must pass)

| Facts | Beginner | A11y | UX | Perf | Sec | Simplicity |
| ----- | -------- | ---- | -- | ---- | --- | ---------- |
| 9     | 9        | 9    | 8  | 8    | 9   | 9          |

## Output

Write `artifacts/<slug>/validation.json` and print a Markdown summary:

```
## Validation report: <slug>
| Dimension | Score | Target | Status | Top finding |
| --------- | ----- | ------ | ------ | ----------- |
...
Blockers:
- ...
Auto-patchable:
- ...
```

## What you must never do

- Never inflate a score to clear a threshold.
- Never hide blockers.
- If a reviewer cannot run (e.g. perf without a build), mark
  `confidence: low` and explain why.