financial-news-video-agent

name	financial-news-video-agent
description	Create proof-first financial news recap videos using article proof, screenshots, charts, and selected VideoDB clips. Use when asked to build a daily market-news video, a proof-backed financial recap, or a news-to-video workflow.

Financial News Video Agent Skill

Use this skill when the user wants to:

• make a daily financial news summary video
• turn market news into a proof-backed recap
• combine charts, screenshots, and video clips into a coherent financial brief
• reuse cached clips and audio instead of recomputing the whole pipeline

Required tools / capabilities

• file read/write/edit
• bash
• Python via uv run
• VideoDB access
• URL/article extraction
• optional screenshot/browser support

Useful external skill references:

• URL-to-Markdown: https://github.com/ash-ishh/skills/tree/main/url-to-markdown
• Playwright: https://github.com/openai/skills/blob/main/skills/.curated/playwright/SKILL.md

Inputs to gather

Before running, gather:

• target date
• market focus (US equities, macro, sectors, etc.)
• target length
• whether the user wants draft or final mode
• whether subtitles are needed

Standard workflow

1. Create date folder

Use:

YYYY-MM-DD/

2. Collect proof sources

Store in:

YYYY-MM-DD/research/

Prefer using a URL-to-Markdown workflow/skill for article extraction.

Collect:

• official releases
• reputable articles
• transcripts where available

3. Capture screenshots

Store in:

YYYY-MM-DD/screenshots/

Prefer a Playwright-based screenshot workflow/skill for webpage capture.

Create video-friendly crops from full screenshots. Do not place raw tall screenshots directly into the final timeline.

4. Build charts

Store in:

assets/YYYY-MM-DD/

At minimum include:

• market overview
• stock-specific chart for each chosen story

5. Search clips with VideoDB

For each chosen story:

• search YouTube/news clips
• upload only the selected source
• index spoken words
• search transcript for the best quote or segment
• save result in clip_registry.json

6. Reuse cached clip/audio assets

Before generating new assets:

• check clip_registry.json
• check audio_registry.json
• reuse existing VideoDB ids whenever possible

7. Assemble the draft video

Include:

• title/opening
• screenshot proof
• chart proof
• actual clip proof when available
• closing summary

8. Subtitle only at the end

Subtitles should be a final pass after draft approval.

Caching policy

Do not rebuild the whole project when one story changes.

If one story changes:

• update only that story's screenshots/crops
• update only that story's selected clip
• update only that story's narration audio if needed
• rebuild the final timeline last

Suggested per-day files

• YYYY-MM-DD/README.md
• YYYY-MM-DD/research/
• YYYY-MM-DD/screenshots/
• YYYY-MM-DD/charts/
• YYYY-MM-DD/reports/
• YYYY-MM-DD/video_build/
• YYYY-MM-DD/scripts/
• YYYY-MM-DD/clip_registry.json
• YYYY-MM-DD/audio_registry.json
• YYYY-MM-DD/render_registry.json
• YYYY-MM-DD/clip_search_notes.md

Deliverables

A good run should produce:

• per-day self-contained example folder
• report markdown
• storyboard markdown
• charts
• screenshots
• final or draft video stream URL
• optional subtitled stream URL

Public example convention

When preparing examples for others, prefer a self-contained layout:

examples/
  YYYY-MM-DD/
    README.md
    research/
    screenshots/
    charts/
    reports/
    scripts/
    video_build/

Reference docs in this repo

• README.md
• docs/SETUP.md
• AGENTS.md