feat(scripts): publish_kaggle + publish_hf + v1 release notes (PR 7.3)

[shaypal5]

Summary

Adds the final two publish scripts and the v1 release runbook. All dry-runs pass locally.

Scripts

scripts/publish_kaggle.py

Three-stage runbook:

1. --dry-run — re-packages release/kaggle/ + lints metadata, no upload
2. Default — uploads as private via kaggle datasets create
3. --public — single-step public upload; or flip via Kaggle web UI after review

Also supports --update MESSAGE for future version bumps (kaggle datasets version).

scripts/publish_hf.py

Same three-stage pattern for HuggingFace:

1. --dry-run — re-packages + lints + runs load_dataset() smoke test (G12.3 / G12.4)
2. Default — uploads as private via huggingface_hub.HfApi.upload_folder
3. --go-public — flips visibility via HfApi.update_repo_visibility

--variant=public|instructor handles both the public (3-tier) and instructor companion repos.

Dry-run results

python scripts/publish_kaggle.py --dry-run
# [ 1/3 ] Packaging release/kaggle/ …  OK
# [ 2/3 ] Linting platform metadata … OK
# Dry-run complete — all pre-flight checks passed.

python scripts/publish_hf.py --dry-run
# [ 1/4 ] Packaging release/huggingface … OK
# [ 2/4 ] Linting platform metadata … OK
# [ 3/4 ] Running load_dataset() smoke tests …
#   OK   config='intro': 3 splits, 5,000 rows total      ← G12.3 ✅
#   OK   config='intermediate': 3 splits, 5,000 rows total
#   OK   config='advanced': 3 splits, 5,000 rows total
# Dry-run complete — all pre-flight checks passed.

python scripts/publish_hf.py --dry-run --variant=instructor
# [ 3/4 ] Running load_dataset() smoke tests …
#   OK   config='intermediate': 3 splits, 5,000 rows total  ← G12.4 ✅
# Dry-run complete — all pre-flight checks passed.

docs/release/v1_release_notes.md

Full release runbook covering:

• 7-step pre-publish checklist (rebuild → validate → dry-runs → ShmuggingFace preview → platform previews)
• Publish steps for all three repos (Kaggle public, HF public, HF instructor)
• Tag + announce instructions
• Change log vs alpha bundles (relational leakage, snapshot leaks, noise clamp, platform hardening)

ShmuggingFace preview

Rebuilt with build_shmuggingface_site.py after PR 8.4 hardening — 48 files generated.

What's left (manual, requires credentials)

• Upload to Kaggle and Hugging Face (see runbook in docs/release/v1_release_notes.md)
• Tag leadforge-lead-scoring-v1
• Announce

---

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds the final two release publish scripts (Kaggle and Hugging Face) plus the v1 release runbook. Each script wraps existing packaging/lint helpers, supports a credential-free --dry-run mode, and (for HF) runs a local load_dataset() smoke test before delegating to the platform CLI/SDK for the actual upload.

Changes:

• New scripts/publish_kaggle.py with package + lint + kaggle datasets create/version flow.
• New scripts/publish_hf.py with package + lint + load_dataset smoke + HfApi.upload_folder, plus a --go-public visibility flip.
• New docs/release/v1_release_notes.md runbook and .agent-plan.md checkbox update.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
scripts/publish_kaggle.py	Three-stage Kaggle publish wrapper around package + lint + Kaggle CLI.
scripts/publish_hf.py	Three-stage HF publish wrapper with load_dataset smoke test and HfApi upload/visibility flip.
docs/release/v1_release_notes.md	v1 release runbook: pre-publish checklist, publish steps, tag/announce, change log.
.agent-plan.md	Marks PR 7.3 done and breaks out remaining manual upload/tag/announce items.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

pr-agent-context report:

This run includes unresolved review comments on PR #87 in repository https://github.com/leadforge-dev/leadforge

For each unresolved review comment, recommend one of: resolve as irrelevant, accept and implement
the recommended solution, open a separate issue and resolve as out-of-scope for this PR, accept and
implement a different solution, or resolve as already treated by the code.

After I reply with my decision per item, implement the accepted actions, resolve the corresponding
PR comments, and push all of these changes in a single commit.

# Copilot Comments

## COPILOT-1
Location: scripts/publish_kaggle.py
URL: https://github.com/leadforge-dev/leadforge/pull/87#discussion_r3310547823
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    The module docstring describes step 3 as `python scripts/publish_kaggle.py --go-public`, but the argparse definition below only exposes `--public` (single-step public create) and there is no `--go-public` flag. The prose right after ("Calls `kaggle datasets metadata --unshare` … actually this is done via the Kaggle web UI or API") also contradicts itself. Please reconcile the docstring with the actual CLI: either remove the `--go-public` section or implement it, and rewrite the step 3 description to match the manual web-UI / `kaggle datasets metadata` + `update` flow that the `main()` epilogue already documents.

## COPILOT-2
Location: scripts/publish_hf.py:284
URL: https://github.com/leadforge-dev/leadforge/pull/87#discussion_r3310547896
Root author: copilot-pull-request-reviewer

Comment:
    `--private` is declared as `action="store_true"` with `default=True`, which means the flag can never be turned off — `args.private` is always `True` regardless of whether the user passes `--private` or not. The help text "(default; review before going public)" and the docstring `--private  Force private upload even when the repo already exists` suggest the intent is a togglable private/public switch, but as written there is no way for a user to upload directly as public on first create (the `--go-public` flow only runs after a separate upload). Consider using `argparse.BooleanOptionalAction` (giving `--private` / `--no-private`) or adding a separate `--public` flag, so the default-private behavior can actually be overridden.

## COPILOT-3
Location: scripts/publish_kaggle.py
URL: https://github.com/leadforge-dev/leadforge/pull/87#discussion_r3310547914
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    The reported visibility treats `args.update` (a new dataset version) as equivalent to "public", but `kaggle datasets version` simply pushes a new version to whatever the repo's current visibility is — it could still be private if the user is iterating before going public. Printing "Upload succeeded (public)." after a version bump can be misleading. Consider only reporting "public" when `args.public` is set, and using a neutral message (e.g. "new version pushed") for the `--update` path.

Run metadata:

Tool ref: v4
Tool version: 4.0.21
Trigger: commit pushed
Workflow run: 26564455345 attempt 1
Comment timestamp: 2026-05-28T08:44:45.007824+00:00
PR head commit: 50263fe00a44d8a28dc2f808c7f1d650d9e19b7b