2.5k
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill video-sourcing- _meta.json637 B
- SKILL.md4.3 KB
Overview
This skill runs a deterministic Video Sourcing Agent to find, compare, and analyze social videos (YouTube, TikTok, Instagram, X) with a compact, predictable chat UX. It auto-bootstraps a pinned runtime when not overridden and returns concise progress and final references suitable for operational workflows. Use it via the /video_sourcing command or by asking for concrete video links and analyses.
How this skill works
When invoked it resolves the user query (stripping the /video_sourcing prefix if present) and launches a pinned runtime to execute a scripted video query. The agent runs in a controlled command pattern, streams NDJSON events, and the runner converts only specific events into chat messages (started, throttled ux_progress, and a terminal event). Default behavior is compact three-message UX; verbose mode is available for debugging and raw payload inspection.
When to use it
- You need concrete video links and short relevance notes for social media research.
- Comparing creator content, trends, or brand mentions across platforms.
- Running reproducible sourcing runs in a deterministic runtime environment.
- Invoking via /video_sourcing to guarantee compact, predictable output.
- Requesting a short list of top results (defaults to top 3).
Best practices
- Prefix queries with /video_sourcing to use the deterministic three-message UX and pinned runtime.
- Keep the query focused (platforms, topic, timeframe) to improve precision.
- Use --event-detail verbose only when you need debugging or raw payloads.
- Expect a fast two-message run when results return quickly; allow progress updates for longer runs.
- Do not ask the agent to fabricate URLs or metrics; it will only return verifiable references.
Example use cases
- /video_sourcing Top brand mentions of BrandX on TikTok last 30 days — returns top 3 videos with one-line relevance notes.
- Ask for creator comparison: "Find recent YouTube explainers on topic Y and highlight best examples."
- Trend check: "Show viral Instagram Reels about hashtag Z this week, top picks only."
- Investigator flow: run in verbose mode to surface payloads when diagnosing missing tooling or env errors.
FAQ
The agent will report the exact missing piece (for example VIDEO_SOURCING_AGENT_ROOT, uv, or a missing API key) and suggest the next action to remediate.
How many videos are returned?
By default the final response lists the top 3 video references (title, url, one-line relevance). If fewer are available, all found results are shown.
How do I get raw/debug output?
Request verbose event detail (--event-detail verbose) or ask for debugging/raw payloads; the agent will switch to verbose mode and stream more details.