Skip to content

Add local YouTube publishing preparation skill - #96

Closed
sfc-gh-cnantasenamat wants to merge 1 commit into
Snowflake-Labs:mainfrom
sfc-gh-cnantasenamat:add-video-publishing-prep
Closed

sfc-gh-cnantasenamat wants to merge 1 commit into
Snowflake-Labs:mainfrom
sfc-gh-cnantasenamat:add-video-publishing-prep

Conversation

@sfc-gh-cnantasenamat

Copy link
Copy Markdown

Summary

Video publishing preparation should deliver usable copy, not just links to output files. This skill prepares transcript-grounded YouTube titles, descriptions, keywords, word-aligned chapters and verified official documentation for existing English-language Snowflake or Streamlit videos. Its final response includes the full publishing content and a linked inventory of generated artifacts.

  • Runs transcription, forced alignment and copy generation locally with FFmpeg, Whisper, WhisperX and a loopback-only Ollama endpoint. Media and transcripts are not sent to cloud inference services, and nothing is published to YouTube.
  • Resolves chapter times from explicit word IDs, retains uncertain alignment flags, and distinguishes timestamp storage precision from acoustic accuracy. Requires semantic review in addition to structural checks.
  • Uses portable paths, optional Python dependencies, bundled scripts/tests and the Snowflake Skills License. Local environments, model weights and video artifacts are excluded.

Example prompt

$video-publishing-prep Prepare YouTube metadata for /absolute/path/tutorial.mp4

Test plan

  • 16 standard-library unit and synthetic workflow tests passed, including exact chapter resolution, rejected flagged anchors, resource rendering, artifact preservation and remote-model rejection.
  • YAML frontmatter and TOML parsed; required contribution fields and title/summary limits checked.
  • Repository validator returned no errors; git diff --check passed.
  • Reviewed for personal paths, credentials, media artifacts and bundled-skill naming conflicts; public duplicate searches found related tools but no close duplicate.
  • Original local workflow exercised on a real video before packaging, with transcription, alignment, local generation, editorial review and resource verification.
  • Fresh-environment full-media run of this portable package. Tests do not prove acoustic accuracy or model-output quality.

Review notes

  • status: Published follows the current CONTRIBUTING.md and existing published skills. The older validator recommends stable/beta/draft and emits a non-blocking warning. Both name/author and id/authors metadata are supplied.
  • English alignment only; models and system executables must be available locally. Package/model installation requires approval. Large transcripts require section-wise generation rather than silent truncation.
  • Pre-PR review was advisory. The contributor confirmed the employee contribution requirements and license choice before submission.

Generated with Snowflake CoCo

Prepare transcript-grounded metadata with reviewed word-level chapter anchors and verified documentation, and require complete publishing content in the final chat response.

Generated with [Snowflake CoCo](https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code)

Co-authored-by: Snowflake CoCo <noreply@snowflake.com>

@snowflake-security-bot snowflake-security-bot Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Snowflake Security Review

Security grade: A — Passed ✅

This PR was classified as LOW risk by the automated pre-screen.

@sfc-gh-cnantasenamat

Copy link
Copy Markdown
Author

Superseded by #98, which includes the renamed tutorial-video-publishing alongside tutorial-video-editor and tutorial-video-thumbnail in one coordinated submission. Closing this standalone PR; branch retained.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant