Video Source Verification

Use this skill when a user gives a specific claim from a video (especially YouTube) and asks for quick verification, evidence extraction, and archival notes.

Quick Start

1. Pull direct evidence first:

node C:/Users/issda/.codex/skills/hydra-node-terminal-browsing/scripts/hydra_terminal_browse.mjs \
  --url "<claim-article-or-video-page-url>" \
  --out "C:/Users/issda/SCBE-AETHERMOORE/docs/research/evidence/<slug>.json"

2. Verify with at least one independent source:

node C:/Users/issda/.codex/skills/hydra-node-terminal-browsing/scripts/hydra_terminal_browse.mjs --url "https://www.toi.com/..." --out "C:/Users/issda/SCBE-AETHERMOORE/docs/research/evidence/<slug>/source_2.json"
node C:/Users/issda/.codex/skills/hydra-node-terminal-browsing/scripts/hydra_terminal_browse.mjs --url "https://www.other-source.com/..." --out "C:/Users/issda/SCBE-AETHERMOORE/docs/research/evidence/<slug>/source_3.json"

3. If claim source is JS-driven (YouTube/watch pages, dynamic embeds), run a Playwright pass:

$env:PWCLI="C:/Users/issda/.codex/skills/playwright/scripts/playwright_cli.sh"
"$env:PWCLI" open <url> --headed
"$env:PWCLI" snapshot
"$env:PWCLI" screenshot

4. Summarize and archive in markdown:

# Save as one deterministic claim note
Set-Content -Path C:/Users/issda/SCBE-AETHERMOORE/docs/research/<slug>.md -Value @"
# Claim: ...
- Source A:
- Source B:
- Evidence quality: low|medium|high
"@

5. Optional: for multi-hop internet-workflow runs, trigger the pipeline tuner after baseline checks:

python C:/Users/issda/.codex/skills/scbe-internet-workflow-synthesis/scripts/run_e2e_pipeline.py --repo-root C:/Users/issda/SCBE-AETHERMOORE --profile training/internet_workflow_profile.json

Evidence Contract (required)

Each claim should emit JSON artifacts with these keys:

• url
• resolved_url
• status
• title
• text_excerpt
• links
• metrics
• fetched_at

Standard Confidence Rubric

• high: original source + direct source of claim + at least one independent corroborating source
• medium: direct claim source + 1 secondary source, but no direct benchmark details
• low: only rumor/aggregate reposting, no direct source or contradictory signals

Safety Rules

• Prefer primary/near-primary sources over re-posts.
• Never store secrets or API keys in research artifacts.
• Keep outputs deterministic and timestamped.
• If claim is impactful, add a short docs/research/<slug>.md note with explicit uncertainty language.

Integration Notes

• For repo-level training workflows, route notable verified claims to:
  • docs/research/
  • docs/notes/ if you need a long-term cross-agent index
• Use this skill when you want reusable, repeatable research steps that can be performed by any local AI agent with your skill registry.