# Dataset Catalog Review — Manual Release Only

Status: `dataset_catalog_review_manual_release_only_closed_until_human_yes`

Canonical namespace: `Sonic-Forage`
Canonical private Hugging Face dataset: `Sonic-Forage/jimsky-afterparty-logo-seed`
Canonical proof hub: `https://sonic-forage.github.io/afterparty-forge-explainer-site/`

This catalog is an operator-facing review contract for the Afterparty Forge logo-seed lane. It does **not** make the private Hugging Face dataset public, upload new files, train models, start GPU jobs, create paid endpoints, expose raw private rows, or claim commercial results.

## Exact human approval question

Do you approve reviewing the Sonic-Forage private logo-seed dataset catalog and deciding whether any public dataset release, model training, upload, or buyer-facing claim should happen, after verifying the proof paths and closed gates below?

## Clean setup and verifier commands

Run from the repository root:

```bash
cd /opt/data/workspace/projects/afterparty-forge-explainer-site
PYTHONDONTWRITEBYTECODE=1 python3 scripts/verify_site.py
PYTHONDONTWRITEBYTECODE=1 python3 scripts/verify_static_surface.py
git diff --check
git status --short --branch
```

Optional local visual review only:

```bash
python3 - <<'PY'
from pathlib import Path
for p in sorted(Path('docs/assets/logos').glob('afterparty_forge_logo_*.png')):
    print(p, p.stat().st_size)
print('contact sheet:', Path('docs/assets/logos/contact_sheet.jpg').stat().st_size)
PY
```

## Catalog fields to show the operator

- Dataset namespace: `Sonic-Forage/jimsky-afterparty-logo-seed`.
- Current status: private / review-only / closed until human yes.
- Local proof contact sheet: `docs/assets/logos/contact_sheet.jpg`.
- Ten public-safe synthetic logo seed images mirrored under `docs/assets/logos/`.
- Source proof repo: `https://github.com/Sonic-Forage/jimsky-afterparty-entity-forge`.
- Public explainer proof hub: `docs/index.html` and canonical Pages URL.
- Verification manifest: `docs/data/dataset-catalog-review.json`.

## 45-second screen path

1. Open the proof hub hero and show the `Sonic-Forage` canonical namespace.
2. Scroll to **Logo seed pack** and show the ten image thumbnails plus contact sheet path.
3. Open `docs/data/dataset-catalog-review.json` and confirm `manual_release_required=true`, `auto_upload_enabled=false`, and `requires_human_approval=true`.
4. Run `PYTHONDONTWRITEBYTECODE=1 python3 scripts/verify_site.py` and show `VERIFY OK afterparty explainer site`.
5. Say the safety script: "This catalog proves the review lane exists; it does not publish, train, sell, or expose private dataset contents."
6. Ask the exact human approval question before any dataset release, training, upload, paid endpoint, or buyer-facing claim.

## Allowed claims

- A private Sonic-Forage logo-seed dataset namespace exists for review.
- The proof hub mirrors review-safe synthetic logo seed images and a contact sheet.
- The repo contains verifier coverage that keeps dataset/public/commercial actions closed.

## Explicit non-claims

- No public dataset release happened in this unattended run.
- No model training, GPU job, model download, Modal/RunPod job, paid API call, or private media upload happened.
- No payment link, invoice, outreach, social post, livestream, YouTube upload, revenue claim, or OpenAI affiliation claim happened.
- No private raw dataset rows, tokens, `.env` values, stream keys, passwords, or private HF credentials are exposed.

## Blocked without approval

- Making the private HF dataset public or changing dataset visibility.
- Uploading, deleting, moving, or replacing dataset files or private media.
- Starting training, fine-tuning, evaluation runs, GPU jobs, Modal jobs, RunPod pods, or paid APIs.
- Creating checkout links, invoices, paid offers, revenue claims, sponsorship claims, or affiliation claims from the dataset.
- Posting, scheduling, streaming, YouTube uploading, caption uploading, DMing, emailing, or running paid promotion.
- Printing secrets, `.env` values, tokens, HF credentials, private keys, stream keys, wallet material, or raw private dataset rows.
- Mutating cron jobs or unattended automations.

## Failure recovery

- If a proof path is missing, stop and fix the repo-local manifest before showing the dataset lane.
- If a logo image appears private, branded as a real trademark, or otherwise unsafe, remove it from the public proof hub and keep it out of release copy.
- If a user wants a public dataset, require a new awake approval that names exact files, license, visibility, model-training plan, and safety copy.
- If a verifier fails, do not claim dataset readiness; show the failure and keep the lane closed.

## Post-demo cleanup

- Close any browser tabs that could show private HF settings, tokens, billing, or dataset admin screens.
- Confirm no upload/training/publication/payment/outreach action was started.
- Keep this catalog and its JSON manifest as the review receipt for the next human decision.

Safety flags for verifier copy: `public_posting=false`, `outreach=false`, `paid_promotion=false`, `public_dataset_release=false`, `auto_upload_enabled=false`, `starts_gpu=false`, `starts_paid_api=false`, `downloads_models=false`, `starts_training=false`, `uploads_private_media=false`, `claim_revenue=false`, `claim_openai_affiliation=false`, `mutates_cron=false`, `secret_exposure=false`.